View
3
Download
0
Category
Preview:
Citation preview
Gravity and Centrifugal Casting of Light Metal
Alloys Using Rapidly Produced Sand Moulds
Nicholas McKenna
A thesis submitted to
Auckland University of Technology
In fulfilment of the requirements for the degree
of
Masters of Engineering
July 2010
School of Engineering
Primary Supervisor: Dr. Sarat Singamneni
i
ABSTRACT
Traditional sand casting is a well understood method to produce metal shapes
and has been used for many years in industry. It is a relatively simple method
but has a significant drawback, with the requirement of a pattern to form the
internal cavity. Patterns are produced at high cost through Computer
Numerical Controlled machining or wood pattern making with significantly high
lead times.
Rapid Prototyping is seen as a solution to this problem, with the ability to
produce sand moulds directly from Computer Aided Design platforms and
thus eliminate the requirement of a pattern. Through layered manufacturing,
the sand mould can be produced with complex internal geometry directly,
minimising labour costs and involving short waiting times. While initial
research was mainly concerned with the use of Selective Laser Sintering, with
the advent of 3D printing, pattern-less sand moulds can be produced more
easily and cheaply. With the process gaining more and more popularity, there
was a need to scientifically assess the suitability of the process for sand
casting as well as, establish influences of typical process parameters on
significant responses.
Critical mould properties, such as permeability and compressive strength,
were investigated with respect to varying time and temperature of baking. To
this end, mathematical models of permeability and compressive strength were
developed. Also, the influence of mould material, mould coating, alloy type
and pouring temperatures were investigated in static sand casting of light
metals. Further work utilised the centrifugal casting process using these 3D
printed moulds to establish links between process factors, such as rotational
speed and cast strength using light metals.
Compressive strength results for the rapidly produced materials were
acceptable compared to traditional values. Permeability was however lower
ii
than commonly used foundry sand. Results showed, nevertheless, that
permeability and compressive strength were both improved by baking times
and temperatures. Significant model effects were established for ZP131 and
ZCast501 with respect to increased compressive strength and mould
permeability.
Multi-factorial experiments involving simultaneous variation of factors such as
mould materials, surface coatings, alloys and pouring temperatures were
conducted and static casting results in general show good as-cast mechanical
properties with the factors having significant effects on surface roughness,
percent elongation and hardness.
Centrifugal casting of aluminium alloys initially produced below average
tensile properties, due to the large presence of hydrogen porosity. However,
upon degassing, much improved tensile strengths were obtained, being
superior to both static casting and traditionally sand cast aluminium. Also a
Magnesium alloy was successfully trialled with the centrifugal process using
3D printed moulds in spite of numerous practical difficulties.
Substantial data relating to the process factors for mould materials and
casting processes was produced. Analysis of factor influences facilitated
optimum process configurations for the production of moulds and castings.
These combinations of factors at optimum levels comprehensively showed
that light metals such as aluminium and magnesium alloys could be
successfully processed by rapidly produced moulds, both statically and
centrifugally.
iii
ACKNOWLEDGEMENTS
I would like to express my profound thanks to my primary supervisor, Dr Sarat
Singamneni, senior lecturer, for his invaluable help and guidance through
every stage of this project.
I would also like to thank my secondary supervisor, Prof. Darius Singh, for his
overall input and commitment in helping organise the logistics for undertaking
this work.
I also wish to note special thanks to the local supporting company,
Centracast, for kindly allowing generous use of their casting facilities and
personnel.
This project was not possible without funding from the Foundation for
Research, Science and Technology, which kindly granted a scholarship to the
author (contract number CCWE0901), and allowed this work to be
undertaken.
Finally, I would also like to sincerely thank my partner Amie, mother Susan
and father Colin for their continuous support, love, help, understanding and
advice during the duration of this project.
iv
TABLE OF CONTENTS
ABSTRACT .................................................................................................... i
ACKNOWLEDGEMENTS ............................................................................. iii
TABLE OF CONTENTS ................................................................................ iv
AUTHORSHIP ............................................................................................ viii
LIST OF TABLES ......................................................................................... ix
LIST OF FIGURES ........................................................................................ xi
ABBRIEVIATIONS ..................................................................................... xvii
NOMENCLATURE .................................................................................... xviii
CHAPTER 1 INTRODUCTION...................................................................... 1
1.1 Rapid Prototyping...................................................................... 1
1.1.1 Evolution of RP towards Rapid Manufacturing .............. 6
1.2 Rapid Prototyping and Casting .................................................. 8
1.2.1 Rapid Casting ............................................................... 9
1.3 Literature Review .................................................................... 11
1.4 Research Questions, Hypotheses and Project Objectives ....... 35
1.5 Methodology ........................................................................... 37
CHAPTER 2 CHARACTERISTICS OF 3D PRINTED MOULDS ................. 39
2.1 Mould Materials ....................................................................... 39
2.1.1 Basic Ingredients of the Mould Materials ..................... 39
2.2 Experimental Design for Mould Material Analysis .................... 42
2.2.1 ANOVA Calculations ................................................... 46
v
2.3 Experimental Methodology for Testing Mould Materials .......... 50
2.3.1 Mould Compressive Strength (MCS) ........................... 51
2.3.2 Mould Permeability (MP) ............................................. 52
2.4 ZP131 Material Results and Discussion .................................. 55
2.4.1 MCS Model ................................................................. 55
2.4.2 MP Model .................................................................... 59
2.5 ZCast501 ................................................................................ 62
2.5.1 ZCast MCS ................................................................. 62
2.5.2 ZCast MP .................................................................... 66
2.5.3 A Comparison Between the Two Materials .................. 68
CHAPTER 3 CHARACTERISTICS OF STATIC CASTINGS PRODUCED IN
3D PRINTED MOULDS................................................................................ 71
3.1 Casting in Printed Moulds ....................................................... 71
3.2 Experimental Plan ................................................................... 73
3.2.1 Taguchi Response and ANOVA .................................. 76
3.2.2 Pooling of Factors ....................................................... 77
3.3 Mould Design .......................................................................... 77
3.4 Static Casting Experiments ..................................................... 79
3.4.1 Mechanical Testing of Castings ................................... 80
3.5 Static Casting Results and Discussions .................................. 83
3.5.1 Surface Roughness ..................................................... 83
3.5.2 Ultimate Tensile Strength (UTS) .................................. 89
3.5.3 Percent Elongation ...................................................... 94
3.5.4 Brinell Hardness .......................................................... 97
3.6 Metallographic Analysis......................................................... 100
3.6.1 ASTM Grain Number ................................................. 100
3.6.2 Magnesium alloy: AZ91 ............................................. 101
3.6.3 Magnesium alloy: AM-SC1 ........................................ 109
3.6.4 Aluminium: A356 ....................................................... 111
3.7 Summary .............................................................................. 121
vi
CHAPTER 4 CENTRIFGUAL CASTING IN RP MOULDS ........................ 122
4.1 Centrifugal Casting ................................................................ 122
4.2 Background ........................................................................... 123
4.3 Analytical Model .................................................................... 125
4.4 Results from the Analytical Model ......................................... 135
4.5 Centrifugal Casting Experimental Design and Setup ............. 137
4.5.1 Mould Design ............................................................ 140
4.5.2 Setup and Methodology ............................................ 140
4.5.3 Experimental Procedure ............................................ 142
4.6 Centrifugal Casting Results ................................................... 143
4.6.1 Surface Roughness ................................................... 145
4.6.2 Yield Strength ........................................................... 153
4.6.3 Ultimate Tensile Strength (UTS) ................................ 158
4.6.4 Percent Elongation .................................................... 164
4.7 Macro and Micro Structural Examination ............................... 167
4.7.1 Macrostructures ........................................................ 167
4.7.2 Microstructures.......................................................... 169
4.7.3 Fractography of Centrifugal Castings ........................ 175
4.8 Further Centrifugal Casting ................................................... 178
4.8.1 An Improvised Centrifugal Casting Setup .................. 178
4.8.2 Centrifugal Casting Trials with the New Setup ........... 179
4.8.3 Results of the Modified Centrifugal Casting Trials ..... 180
4.9 Summary of Centrifugal Casting Trials .................................. 186
CHAPTER 5 CONCLUSIONS .................................................................. 188
REFERENCES ........................................................................................... 191
APPENDIX A MOULD MATERIAL ........................................................... A-1
A.1 Experimental Design .............................................................. A-1
A.2 Permeability and Compressive Stress Calculations................ A-4
A.3 Raw Experimental Results From Material Testing .................. A-5
vii
A.3.1 Permeability Data ....................................................... A-6
A.3.2 Compressive Strength Data ....................................... A-8
APPENDIX B STATIC CASTING INFORMATION ................................... B-1
B.1 Design of Experiments and ANOVA ....................................... B-1
B.1.1 Sum of squares .......................................................... B-2
B.2 Mechanical testing data.......................................................... B-4
B.2.1 ASTM grain size ......................................................... B-6
B.3 Microstructure Evaluation ....................................................... B-9
B.4 Castings Produced in ZP131 Moulds ................................... B-12
B.4.1 Castings Produced in ZCast Moulds ........................ B-15
B.5 Tensile Part Setup................................................................ B-18
B.6 Static Mould Design ............................................................. B-18
B.7 Alloy Constitutes .................................................................. B-19
APPENDIX C CENTRIFUGAL CASTING DATA ...................................... C-1
C.1 Experimental Design ............................................................. C-1
C.2 Mechanical Testing Data ....................................................... C-3
C.3 Etchants Used for Cast Alloys ............................................... C-5
C.4 As-Cast Macrostructures ....................................................... C-7
C.5 SEM Photographs ............................................................... C-10
C.6 Die and Mould Design Drawings ......................................... C-13
C.7 Mould Design ...................................................................... C-17
C.8 Tensile testing part .............................................................. C-18
C.9 Publications......................................................................... C-19
viii
AUTHORSHIP
“I hereby declare that this submission is my own work and that, to the best of
my knowledge and belief, it contains no material previously published or
written by another person (except where explicitly defined in the
acknowledgments), nor material which to a substantial extent has been
submitted for the award of any other degree or diploma of a university or other
institution of higher learning”
Signed: Nicholas McKenna Date: 30th July 2010
ix
LIST OF TABLES
Table 2.1 The ZB60 binder ingredients for the ZP131 material [49]. ............. 42
Table 2.2 ZB58 binder composition for the ZCast501 material [50]. ............. 42
Table 2.3 The Factorial part of CCD ............................................................. 44
Table 2.4 Experimental design table showing (a) the x matrix of the time and
temperatures as coded variables and (b) showing the design table
with the natural and coded variable combinations ....................... 45
Table 2.5 ANOVA of compressive strength model of ZP131 ........................ 56
Table 2.6 ANOVA of permeability model of ZP131 ...................................... 59
Table 2.7 ANOVA table of the MCS model ................................................... 63
Table 2.8 The ANOVA table of the permeability model ................................. 66
Table 3.1 Taguchi L9 experimental design table with natural variables ........ 75
Table 3.2 Factor and level combinations for static casting trials ................... 75
Table 3.3 ANOVA table of the surface roughness response ......................... 84
Table 3.4 Static L9 experimental design table .............................................. 88
Table 3.5 Overall ranking of factors and their level from static casting trials . 88
Table 3.6 ANOVA of UTS response ............................................................. 89
Table 3.7 ANOVA table for percent elongation response model ................... 95
Table 3.8 ANOVA of Brinell hardness response ........................................... 97
Table 4.1 Values of G developed at different speeds with respect to different
radii ........................................................................................... 137
Table 4.2 Summarised factor and level combinations for the centrifugal
casting trials .............................................................................. 139
Table 4.3 Centrifugal casting experimental design table ............................. 139
Table 4.4 Overall results of mechanical testing of centrifugal test bars ....... 144
Table 4.5 Contribution and significance of individual terms for the surface
response model ......................................................................... 152
Table 4.6 ANOVA table showing linear, square and interaction model effects
.................................................................................................. 152
Table 4.7 Breakdown of model parameters in the yield strength model ...... 157
Table 4.8 ANOVA table of the yield strength model response .................... 157
Table 4.9 Results of T-Tests conducted between each speed range. ......... 158
x
Table 4.10 Tensile properties of centrifugally cast degassed specimens .... 160
Table 4.11 Contribution of individual model parameters in the UTS model . 162
Table 4.12 ANOVA of the UTS model response ......................................... 163
Table 4.13 Significance of the individual model parameters in the percent
elongation model ....................................................................... 166
Table 4.14 ANOVA of the various effects present in the percent elongation
model. ....................................................................................... 166
Table 4.15 Tensile testing results from centrifugal castings, produced with
CC601 aluminium ...................................................................... 180
Table 4.16 Results of centrifugal magnesium castings ............................... 184
xi
LIST OF FIGURES
Figure 1.1 The Depiction of the SL process [3] ............................................... 2
Figure 1.2 The FDM process [4] ..................................................................... 3
Figure 1.3 SLS machine process [5] ............................................................... 4
Figure 1.4 LOM process showing passing build sheet and build platform [4] .. 5
Figure 1.5 Schematic of 3D printing process showing piston and binder
application [6] ................................................................................ 5
Figure 1.6 AUT‟s ZCorporation Z310 3D printer ............................................. 6
Figure 1.7 SolidWorks computer model showing the internal cavity of a test
mould ............................................................................................ 8
Figure 2.1 Spectra of ZP131 mould material by X-ray diffraction .................. 40
Figure 2.2 The XRD spectra graph showing the refraction angles of the ZCast
501 material ................................................................................ 41
Figure 2.3 Data plot of the points showing the different sampling combinations
of the central composite experimental design. ............................. 43
Figure 2.4 Photograph of the desiccator used to prevent moisture formation
.................................................................................................... 51
Figure 2.5 Photograph depicting the compressive test of the mould material.
.................................................................................................... 52
Figure 2.6 The Permeability set up showing (a) the sealant used to adhere the
membrane to the specimens (b) rubber rings and specimen with
membrane applied and (c) load cell apparatus ready for testing .. 53
Figure 2.7 Photograph of the Tri-axial cell set up for the permeability testing 54
Figure 2.8 Close up of tri-axial cell test apparatus whilst testing ................... 55
Figure 2.9 Compressive strength Vs baking time and temperature for ZP131
.................................................................................................... 57
Figure 2.10 The Variation of Mould Compressive Strength of ZP131 with (a)
baking temperature and (b) baking time ...................................... 58
Figure 2.11 Permeability Vs baking time and temperature in the case of
ZP131 ......................................................................................... 60
xii
Figure 2.12 Permeability plots with respect to (a) varying temperature and (b)
time ............................................................................................. 61
Figure 2.13 Mould Compressive Strength Vs time and temperature of baking
for ZCast 501 .............................................................................. 63
Figure 2.14 ZCast501 mould sample baking temperature was 150°C and time
equal to 4 hours........................................................................... 64
Figure 2.15 SEM photograph showing the embrittled gypsum at the grain
boundary ..................................................................................... 65
Figure 2.16 Permeability Vs baking time and temperature in the case of
ZCast501..................................................................................... 67
Figure 2.17 SEM photographs showing the gypsum structure of a mould
baked at (a) 150°C and 4 hours and (b) modified structure of a
mould baked at 270°C for 6 hours ............................................... 68
Figure 3.1 CAD drawings of the initial static mould design, showing (a) Bottom
half of the mould and (b) 3D view of completed mould design. .... 78
Figure 3.2 Sectional CAD drawing initial mould design, showing pouring
entrance cup, cylindrical cavity and riser ..................................... 79
Figure 3.3 Photographs showing AUT University induction furnace setup (left)
during casting of Mg alloy (right) .................................................. 80
Figure 3.4 Degassing lance used in casting trials ......................................... 80
Figure 3.5 Taylor Hobson Talysur50 surface testing machine ...................... 81
Figure 3.6 Photographs of the tensile testing performed at AUT with (a)
showing the testing machine set up and (b) showing close up the
stressing of a tensile specimen .................................................... 82
Figure 3.7 Brinell Hardness testing apparatus .............................................. 82
Figure 3.8 S/N ratio of mould material in the surface roughness results ....... 84
Figure 3.9 The SEM close up photograph of an angular Olivine sand grain
present in the material ................................................................. 85
Figure 3.10 S/N ratios of (a) Mould coating and (b) Alloy type levels in the
surface roughness response. ...................................................... 86
Figure 3.11 S/N ratio of the pouring temperature factor levels on the surface
roughness values ........................................................................ 87
Figure 3.12 S/N ratio of the Alloy type factor levels in the UTS response ..... 90
xiii
Figure 3.13 S/N ratio of the Pouring temperature Factor levels in relation to
the UTS response. ...................................................................... 90
Figure 3.14 S/N ratio of the mould coating factor levels in relation to the UTS
response. .................................................................................... 91
Figure 3.15 S/N ratio of the mould material factor levels in the UTS response
.................................................................................................... 92
Figure 3.16 S/N ratios relating to the percent elongation values at factors (a)
Alloy type (b) Mould material (c) Mould coating and (d) pouring
temperature ................................................................................. 95
Figure 3.17 S/N ratio of the alloy type factor levels in the Brinell hardness
response. .................................................................................... 98
Figure 3.18 S/N ratio of the Mould material factor levels in the Brinell
hardness response. ..................................................................... 98
Figure 3.19 S/N ratios relating to the Brinell hardness values at factors (a)
Mould coating and (b) pouring temperature. ................................ 99
Figure 3.20 Average ASTM grain number produced in different moulds ..... 101
Figure 3.21 Magnesium-Aluminium binary phase diagram [78] .................. 103
Figure 3.22 AZ91 microstructures in (a&b) ZP131 (c&d) ZCast and (e&f) Silica
foundry moulds .......................................................................... 104
Figure 3.23 Micrographs showing (a) partially divorced and (b) fully divorced
eutectic structures in AZ91 magnesium alloy ............................. 106
Figure 3.24 Photomicrograph of AZ91 casting showing areas of micro-voids
(X50 magnification) ................................................................... 107
Figure 3.25 SEM photographs showing porosity present on the fractured
surfaces of the AZ91 Castings (X130 (left) and X500 (right)
magnification respectively). ....................................................... 108
Figure 3.26 SEM fracture surface showing quasi-cleavage fracture surface in
(a) current work and (b) Literature ............................................. 108
Figure 3.27 Photomicrographs of SC1 produced in (a&b) in ZP131 and (c&d)
ZCast501 and (e&f) Silica foundry sand moulds ........................ 110
Figure 3.28 SEM photographs showing fracture surface on the primary (a) Mg
dendrites and (b) more ductile regions on the same sample
(magnifications are X500 and X4000 respectively). ................... 111
xiv
Figure 3.29 Microphotographs showing A356 cast structures in (a&b) ZP131
(b&c) ZCast501 and (d&e) Silica foundry sand .......................... 113
Figure 3.30 Al-Si alloy, showing (a) fine modified eutectic silicon and (b)
coarse eutectic silicon. .............................................................. 114
Figure 3.31 Photomicrograph showing plate like Fe intermetallics in A356
casting produced in 3D printed moulds ...................................... 116
Figure 3.32 Un-degassed showing typical hydrogen porosity in (a)
microstructure and (b) macrostructure produced in RP moulds . 117
Figure 3.33 SEM photographs showing large pores from su8spected
hydrogen porosity ...................................................................... 119
Figure 3.34 SEM fracture surfaces showing (a) typical dimpled, ductile
fracture surface [93] and (b) a sample from the current research ,
exhibiting a lack of a rim around each dimple. ........................... 120
Figure 3.35 Different fracture modes, with arrows showing the direction of the
applied loads [93] ...................................................................... 120
Figure 3.36 A Crack path seen to initiate from pores due to suspected
hydrogen porosity. ..................................................................... 121
Figure 4.1 Vertical semi-centrifugal casting setup [95] ................................ 123
Figure 4.2 Illustration of the centrifuge technique with multiple castings placed
around the central sprue [95] ..................................................... 124
Figure 4.3 Schematic of the forces acting on an element of liquid under the
absence of centripetal forces, at constant angular velocity ........ 126
Figure 4.4 Results of the analytical model showing (a) the velocity at different
radii from the centre of rotation and (b) fluid pressures developed
at the same radii. ....................................................................... 136
Figure 4.5 CAD models of the centrifugal ZP131 plaster moulds used in the
experimental trials, showing (a) samples angles 90°to runner and
(b) samples orientated at 0 and 45° in relation to the runner bar.
.................................................................................................. 140
Figure 4.6 CAD sectional model of the centrifugal die and mould setup ..... 141
Figure 4.7 Centrifugal casting setup showing steel die, pouring cup and
holding crucible ......................................................................... 142
xv
Figure 4.8 Graphical form of overall results from testing at various rotational
speeds, specifically (a) surface roughness (b) ASTM grain size of
primary dendrites and (c) tensile properties. .............................. 145
Figure 4.9 Possible influence of pressure on surface roughness at (a) low &
(b) high pressures ..................................................................... 147
Figure 4.10 Typical cast surfaces at (a) 150 RPM and (b) 450 rpm ............ 148
Figure 4.11 Graphical plots of surface roughness model results at runner-
cavity (r/c) ratios of (a) 0.8:1 (b) 1:1 (c) 1.1:1 and (d) 1.2:1. ....... 151
Figure 4.12 Yield strength variation with rotational speed ........................... 154
Figure 4.13 Comparison between (a) Static casting and (b) Centrifugal
castings produced in ZP131 plaster moulds (both X50) ............. 155
Figure 4.14 The Graphical plots of yield strength model at r/c ratios of (a) 0.8
(b) 1 (c) 1.1 and (d) 1.2.............................................................. 156
Figure 4.15 The Graphical plots of Ultimate Tensile strength model at r/c
ratios of (a) 0.8:1 (b) 1:1 (c) 1.1:1 and (d) 1.2:1. ........................ 161
Figure 4.16 UTS values of specimens ignoring results of trial 15................ 162
Figure 4.17 Individual plots of rotational speed and cavity angle with respect
to the percent elongation model at r/c ratios of (a) 0.8:1 (b) 1:1 (c)
1.1:1 and (d)1.2:1. ..................................................................... 165
Figure 4.18 Macrostructures of cast specimens at (a) 150 (b) 300 and (c) 450
RPM .......................................................................................... 168
Figure 4.19 Microstructures of centrifugal castings, spun at 150RPM at X100
and X400 magnifications respectively (all) for (a & b) Trial 1 – 45°,
1:1 runner ratio (c & d) Trial 7 – 90°, 1.2:1 runner ratio and (d & e)
Trial 3 – 180°, 1:1 runner ratio. .................................................. 170
Figure 4.20 Microstructures of castings spun at 300RPM at conditions (a & b)
Trial 11 – 45°, 1.2:1 runner ratio (c & d) Trial 15 – 90°, 1:1 runner
ratio and (d & e) Trial 12 – 90°, 1:1 runner ratio. ........................ 171
Figure 4.21 The Microstructures of castings spun at 450RPM at conditions (a
& b) Trial 2 – 45°, 1:1 runner ratio (c & d) Trial 8 – 90°, 1.2:1 runner
ratio and (d & e) Trial 4 – 180°, 1:1 runner ratio. ........................ 172
Figure 4.22 Microporosity present at the three different rotational speeds,
namely (a) 150RPM (b) 300RPM and (c) 450RPM .................... 173
xvi
Figure 4.23 Photomicrographs of entrained oxide films in castings conducted
at (a) 150RPM (b) 300RPM and (c) 450RPM (all X50
magnification) ............................................................................ 174
Figure 4.24 Die and machine setup, with fall height depicted ..................... 175
Figure 4.25 SEM photographs of the fractured surface at (a) 150 (b) 300 and
(c) 450RPM. .............................................................................. 176
Figure 4.26 SEM photographs showing (a) long oxide film and (b) Fe-
intermetallic particle ................................................................... 177
Figure 4.27 Improvised potter‟s wheel used for CC of Mg .......................... 178
Figure 4.28 Al as-cast microstructure of lance degassed centrifugal castings
at (a) X50 (b) X100 and (c) X400 ............................................... 182
Figure 4.29 AZ91 CC shown in ZP131 printed mould (left) with close-up of
cast surface shown on the right ................................................. 183
Figure 4.30 Photomicrographs of centrifugally cast AZ91 in ZP131 moulds at
(a) X50 (b) X100 and (c) X400................................................... 185
Figure 4.31 Suspected shrinkage porosity in AZ91 CC (X100) ................... 186
xvii
ABBRIEVIATIONS
RP Rapid Prototyping
CAD Computer Aided Design
AM Additive Manufacturing
SL Sterolithography
FDM Fused Deposition Modelling
SLS Selective Laser Sintering
3D printing Three Dimensional Printing
LOM Laminated Object Manufacturing
UV Ultraviolet
ABS Acrylonitrile Butadiene Styrene
D Dimensional
IC Investment Casting
CNC Computer Numerically Controlled
CMM Co-ordinate Measurement Machine
SEM Scanning Electron Microscope
ZCorp ZCorporation
TDS Thermal Distortion Testing
IT International Tolerance
HCP Hexagonal Closed Package
SDAS Secondary Dendrite Arm Spacing
CAST CAST CRC
OA Orthogonal Array
S/N Signal to Noise Ratio
MSD Mean Squared Deviation
FCC Face Centred Cubic
Seq Sequential
Adj Adjusted
T Value from T-Test
xviii
NOMENCLATURE
T Temperature °C
t Time hrs
L Length m
P Pressure Pa
P value Probability value %
Q Flow Rate m3/s
Ra Average surface roughness μm
UTS Ultimate Tensile Strength MPa
A Cross-sectional area m2
F Fisher F value
MCS Mould Compressive Strength MPa
MP Mould Permeability mD
HB Brinell Hardness
σYield Yield Stress MPa
1
CHAPTER 1 INTRODUCTION
1.1 Rapid Prototyping
Rapid prototyping (RP) is a relatively new methodology established in the late
1980‟s, by which material parts can be constructed directly from Computer
Aided Design (CAD) files. Essentially this is accomplished by a virtual slicing
of continuing cross-sections of the parts, thus creating a multi-layer model of
any given part. This continued repetition of the slicing leads to the final shape
of the part being reproduced. This process is also known as Additive
Manufacturing (AM).
Currently, parts are produced in a range of many different materials. These
materials include certain polymers, plaster, starch, foundry sand, powdered
metals, wax and ceramics. Each of these materials can be combined with
various, specific processes, each possessing their particular processing
principles. Nevertheless, the essential layered manufacturing steps remain
more or less the same. The most commonly used RP systems include
Sterolithography (SL), Fused Deposition Modelling (FDM), Selective Laser
Sintering (SLS), Three Dimensional Printing (3D printing) and Laminated
Object Manufacturing (LOM). An initial description of the operation and
capability of these common RP systems first follows first.
Sterolithography (SL)
In 1986, the SL process was patented and the first RP technique was born [1].
This process, shown below in Figure 1.1, uses a photosensitive liquid resin to
form a solid polymer when ultraviolet (UV) light is applied onto the resin
surface [2]. This reaction to the UV light takes place near the surface, creating
solid three dimensional pixels [2]. The UV light is directed by a laser over
each individual layer from information contained in the solid CAD. Once the
layer has been competed the platform containing the resin is lowered to apply
the resin thoroughly and then raised back up so that exactly one layer
2
thickness remains above the scanned surface [2]. The part is then lowered by
one layer after the liquid has settled so that the next layer can be scanned
and so on.
Figure 1.1 The Depiction of the SL process [3]
Fused Deposition modelling (FDM)
The FDM process is an extrusion technique where molten material, usually
Acrylonitrile Butadiene Styrene (ABS) plastic, is passed through a heated
nozzle. The nozzle moves in two dimensions (x, y directions) to print the
individual layer, whilst the build platform is moved in the third direction (Z
direction) after each layer is printed. The material is usually heated near or
just above the melting point and once it is deposited the material „cold welds‟
[2] itself to the previous layer. Unlike the SL process, FDM machines have to
build a support material to hold the deposited material in place. This is usually
by way of a separate nozzle which deposits a relatively cheaper material
which can be dissolved or broken away from the part [2]. Figure 1.2 below
shows the operation of a FDM machine with both the build material nozzle
and the support material nozzle.
3
Figure 1.2 The FDM process [4]
Selective Laser Sintering (SLS)
The SLS process typically uses a high powered CO2 laser to fuse particles
together. The process was originally patented by the University of Texas with
the DTM Corporation commercialising the process in 1992. Figure 1.3 below
shows the overall process setup in which, the particles are fused together by
a laser heating them above their melting point and allowing the particle liquid
phases to mix and fuse [2]. To prevent any thermal distortion, the bed of the
SLS machine is heated to just below the melting point of the material [2]. The
operation of the SLS process is similar to the SL process, with the un-sintered
material acting as the support structure. After each 2D layer is sintered, the
platform is lowered one layer, for the next layer to be sintered. SLS machines
are capable of sintering many different materials such as nylon, polystyrene,
metal powders and suitably coated foundry sand.
4
Figure 1.3 SLS machine process [5]
Although able to sinter most materials, SLS machines are often relatively
expensive and slow to produce parts when compared to 3D printing. Further,
SLS machines tend to use much more energy to operate and powdered
metals suitable for sintering are often expensive.
Laminated Object Manufacturing (LOM)
LOM uses a slightly different technique whereby solid sheets construct the
layered model. The build material is actually a roll of material which is bonded
layer over layer. Figure 1.4 below illustrates the process, where a hot roller
activates a heat sensitive adhesive to fuse the layers together [2]. The shape
of each layer is cut out by a laser, with a careful control over the layer
thickness [2]. The LOM process has several advantages such as relatively
cheap materials and the ability to produce at a much faster speed than other
RP machines. However, this process does produce large amounts of waste
and suffers from an inability to create hollow parts.
5
Figure 1.4 LOM process showing passing build sheet and build platform [4]
Three Dimensional Printing (3D printing)
The 3D printing process consists of a liquid binder that is printed on each
layer, using a standard ink jet print head (based on ZCorporation 3D printers).
Figure 1.5 shows how the process prints each layer in two dimensions,
commonly using a plaster or ceramic powder. Once each layer is printed, the
build platform drops by the thickness of one layer and the feed piston rises by
the same thickness. The roller sweeps across the build platform and
distributes a fresh layer of powder on the build platform. These steps are
repeated until the part is complete.
Figure 1.5 Schematic of 3D printing process showing piston and binder application [6]
6
Figure 1.6 AUT’s ZCorporation Z310 3D printer
Figure 1.6 gives a closer look at the actual arrangement of different
components of the ZPrinter 310 produced by the ZCorporation. This is the
machine used at AUT. The figure shows a standard print head which is
capable of depositing a liquid binder on both powder materials such plaster
(ZP131) or special casting materials (ZCast501). The part is built in layers of
0.1mm, at a speed 20mm per hour, with a maximum build volume of
203X254X203 (all mm).
1.1.1 Evolution of RP towards Rapid Manufacturing
The establishment of the first RP systems manufacturer, 3D Systems Limited
in 1987 was quickly followed by other notable companies such as Stratasys,
EOS, ZCorporation, Pro Metal and DTM between 1988 and 1992 [1]. The
ability to create complex parts quickly and easily from CAD files made this a
popular technology. Since then, these techniques have become valuable tools
for shortening and improving the product design process and avoiding costly
mistakes in pre-production processes. They are also now an inherent feature
of concurrent engineering processes.
7
The potential of RP techniques has been realised in many engineering
disciplines, and it is likely that these techniques will continue to be integrated
into comprehensive manufacturing processes. The ability to increase part
complexity without necessarily increasing product lead times and cost has
now allowed design engineers much greater freedom in product design [7].
This allows the customisation of parts without undue constraint in
manufacturing. The benefits in cost reduction, optimal design and verification
of tooling are all available before full scale manufacturing commences. Profit
margins may increase due to lower fixed operating costs [7] and, equally,
labour resources in manufacturing are also reduced or directed elsewhere.
Essentially, the automatic production of the parts by this technology saves
labour, time and inspection and enhances quality control.
Considering all these advantages, the interest in RP processes has quickly
grown from being used to just produce prototype models with inferior
materials, to rapid tooling applications for direct production of end use parts
using extensive engineering materials. This approach is referred to Rapid
Manufacturing [7]. By definition, rapid manufacturing means the production of
end use parts direct from CAD files, without the need for any complex tooling,
and by means of one or more of the RP technologies. While processes such
as FDM are already competing with traditional injection moulding selective
laser sintering of ceramic and metal powders to produce complex 3D objects
is opening up new areas of application.
Rapid manufacturing however, is also attempted by indirect means, using RP
techniques together with some traditional processes, aiming at an overall
reduction in the total manufacturing lead time. Use of polymer parts produced
by one of the RP techniques as patterns for the Investment Casting (IC)
process is one of the early developments that allowed reducing considerable
time savings in the making of complex patterns. Further use of RP patterns to
save time in production of dies and other tooling has been successfully
applied in many cases. SLS and 3D printing technology has also been applied
to the production of pattern-less sand moulds directly from CAD files for the
casting of light metals.
8
1.2 Rapid Prototyping and Casting
The ability to produce complex shapes and patterns directly from CAD files as
a method to reduce costly and time consuming traditional pattern making
processes stimulated interest from the casting industry. The production lead
time in terms of the fabrication of traditionally made mould patterns by
Computer Numerically Controlled (CNC) machining or woodworking is
typically weeks or months, depending on the complexity of the part to be
produced. The tensile testing mould cavity (pattern), shown below in Figure
1.7, was sent to a local pattern maker and was quoted at $NZ1800 (2009) for
the pattern alone, which was said to take forty hours to complete. This pattern
was relatively basic as the mould comprises of a couple of dog bone shaped
tensile test bars and the sprue and riser system. The pattern making process
alone takes one week and typically work is backed up for a number of weeks,
to have the pattern produced, and from there the mould still has to be created
and then metal cast.
Figure 1.7 SolidWorks computer model showing the internal cavity of a test
mould
Utilising CNC machinery, and a quote from a local CNC machinist, completion
took up to 2 to 3 months in busy periods, down to about 2-3 weeks in quieter
periods to fabricate this pattern. The high demand for the use of CNC means
that lead times can also be highly variable and this could lead to problems if a
part has to be made within a specific time frame. Moreover, the lead times of
9
the mould maker who constructs the mould has also to be considered. So,
typical times to produce the cast part, in terms of traditional pattern making
range from about 3 weeks for wood pattern and anywhere from 3 weeks to 3
months if CNC machining is utilised.
1.2.1 Rapid Casting
RP processes have found wide application in the casting industry, with the
earliest applications being in the IC process, where RP patterns have been
extensively used to create sacrificial patterns for investment in a ceramic
material. SL, FDM, and SLS are all capable of producing patterns for the IC
process, while LOM finds its application in the rapid production of complex
patterns, for processing in a traditional sand casting way.
The lead times of foundries for producing new parts are significantly
determined by the time in which the pattern can be produced. Furthermore it
is common that once a casting is produced, the customer may wish to alter
the design, requiring pattern modification or an entirely new pattern. The
demand to produce patterns suitable for casting by RP therefore became an
attractive option for foundries, starting with prototype modelling and through to
complex completed moulds. As mentioned above, the RP process has
allowed foundries to create expendable patterns that can be used in the IC
process.
The advantages of RP manufacturing processes have led to an increasing
interest in their application to the metal casting sector. This has resulted in
significant research regarding the suitability of different RP technologies in the
IC process. Thus, work was first directed at determining surface roughness,
pattern burn out, accuracy of cast parts and tolerance ratings of printed
patterns. RP techniques produced patterns in a shorter time than traditional
methods as noted above. However, there remained problems with thermal
expansion causing the ceramic shell moulds to crack [8]. Production of the
actual mould therefore continued to be made by traditional means.
10
To overcome problems with pattern burnout and pattern expansion defects,
RP manufacturers began to create materials especially suited for use in the IC
process. 3D Systems created the Quickcast material to be used on SL
machines. This material comprises a hollow internal structure principally for
creating patterns for IC parts. Stratasys Inc subsequently produces parts
suitable for IC by the FDM process. The FDM parts are now produced in a
wax material which is suitable as a pattern in the IC process. SLS machines
soon followed with the introduction of the CastForm material from DTM [2].
CastForm is a specially designed polystyrene material for IC patterns. The
material features a low ash content and is compatible with traditional foundry
practices [2]. The porous CastForm patterns can then be infiltrated with
traditional foundry wax to produce adequate IC patterns comprising of 45%
polystyrene and 55% wax [2]. DTM also produced the TrueForm SLS
material, which is utilised for producing casting patterns. TrueForm however,
is an acrylic-based powder and is not as effective in terms of residual ash
after pattern burnout as the CastForm material [2]. All major RP systems
outlined above have a material capable of being used in the IC process and
traditional IC has benefited from RP patterns. TrueForm however, is an
acrylic-based powder and is not as effective in terms of residual ash after
pattern burnout as the polystyrene/wax based CastForm material [2]. These
enhancements in RP pattern production for IC, resulted in material
improvement in pattern burnout and degradation.
The next stage of development was for RP processes to extend their
application towards the direct production of moulds, making pattern
production redundant. Subsequent to the work on RP patterns as above, it
was realised that RP techniques such as SLS and 3D printing also have the
ability to construct sacrificial moulds directly from CAD files. These
technologies are able to create complex internal geometries and gating
systems for direct mould production. The 3D printing process for production of
moulds in ceramic materials was first achieved in the early 1990‟s with the
commercial licensing of research generated at the Massachusetts Institute of
Technology (MIT). Soligen marketed the jetting of binder onto ceramic
materials for the construction of moulds for investment casting [1]. Other MIT
11
patents associated with 3D printing were commercialised by the ZCorporation
(ZCorp) and Extrudes Hone‟s ProMetal system 1998 and 1997 respectively
[1].
Despite these advances, the overall effectiveness of these rapidly produced
moulds in terms of quality and application, were largely unknown. Similarly,
little work had been done regarding the combinations of the different variables
involved, such as mould characteristics and the mechanical properties of
various cast metals. Furthermore, little comparative analysis had been
undertaken regarding RP and traditional sand casting techniques. The
following literature review is aimed at establishing what has been done in the
casting sector with RP techniques with special emphasis on the 3D printing
process and rapid casting in the context of light metals, such as Aluminium
(Al) and Magnesium (Mg) alloys.
1.3 Literature Review
The establishment of Rapid Prototyping (RP) technologies in the late 1980‟s
and their evolution since then has both led to new applications and significant
changes in specific processes in the engineering industry. Metal casting is
one of the main targets for the application of RP technology. The ability to
produce prototypes quickly from Computer Aided Design (CAD) programmes
has been a valuable tool for design and foundry engineers in eliminating
design flaws before committing to full scale production. The advantages of
using layered manufacturing in the metal casting sector has lead to production
of reusable and sacrificial patterns to reduce both time of production and
tooling costs for short to medium volumes.
The majority of applications of RP patterns have been in the Investment
Casting (IC) sector for sacrificial wax patterns. These patterns are burnt out
over time to leave the internal pattern. Recent research however has utilised
direct printing of sand and ceramic materials for the fabrication of sacrificial
moulds for metal casting. Three dimensional Printing (3D printing), and
Selective Laser Sintering (SLS) techniques have been at the forefront of
creating moulds for casting. The production of broad research in the
12
application of RP technology to the casting sector is reviewed below. This is
followed by a discussion on specific developments in the application of
magnesium and its alloys in gravity casting. Finally there is an introduction to
the specific characteristics and the practical application of centrifugal casting
process. This is the background to the formulation of the research question
that this report addresses.
Constructing new patterns for sacrificial moulds requires both time and skill.
Patterns such as those used in the IC process often have thin sections with
complex shapes, which require further skill and resources to produce. The RP
process is able to create custom made complex shapes and sections, and its
application to create patterns for IC was an obvious use of the technology.
Early use of RP techniques centred on the production of patterns for sacrificial
use in the IC process. IC moulds are either produced indirectly or directly with
RP technology. Indirect production involves using sacrificial RP patterns as
the wax pattern. This also extends into silicone rubber moulding, where the
RP pattern is moulded into the silicone mould to create a cavity for further
fabrication of wax positives. Direct mould production involves the fabrication
of thin ceramic shells by SLS or 3D printing (Direct Shell Production Method,
DSPM) in which metal is cast. Much of the earlier work used indirect
approaches (i.e. sacrificial patterns), with later work focusing on direct mould
production (sacrificial shell moulds).
Use of RP Patterns
Tromans [9] presented an early overview of the various RP processes.
Foundation work with casting and RP was completed with the SL process,
which produced sacrificial patterns for use in IC process. The sacrificial
pattern was made from a non-engineering plastic, which was coated in wax.
This process was widely used and cast parts possessed a good surface
finish. Pham and Dimov [10] showed that other RP processes were also
suitable for producing sacrificial patterns, being; SLS and FDM.
13
Increasing use of sacrificial RP patterns highlighted problems such as shell
cracking upon burnout and promoted investigation into the suitability of
sacrificial RP pattern by Dickens, et al. [11]. The aim was to determine the
suitability of different RP processes in creating sound IC shells from sacrificial
patterns in the IC process. Accuracy of patterns and castings were also
investigated, together with the cast surface quality. The RP systems applied
were LOM, SLA and FDM, all of which created sacrificial wax patterns directly
from CAD files. SLS process created a porous structure which needed further
sealing with wax before being shelled. To test these RP technologies three
foundries were each given the same set of patterns for a window wiper
mechanism for a German automobile, to convert to castings.
All results reported significant time reductions from the avoidance of
traditional prototypes and small scale tooling costs. These traditional costs
were quoted as being between £1,000- £50,000 (1995) with lead times
varying from 1-16 weeks. This did not include the related high commitment of
staff for new projects and products. However, this did not factor in the costs of
skilled operators using RP equipment and subsequent CAD drafting work. It
was thus unclear whether the time and cost saving of using RP technologies
in the IC process outweighed the extra cost of the RP machine and operator
in large scale production. Results of testing from Co-ordinate Measurement
Machine (CMM) revealed large standard deviations in accuracy with
inadequate tolerances found in both the patterns and the castings. Lower
variability was observed in the surface roughness values, with acceptable
figures ranging from 4-25μm. Foundry experience was evidently a critical
factor in the conversion of RP patterns to sound investment castings. More
recent research has suggested that SLA, SLS and FDM technologies are
suitable RP processes for investment and vacuum casting [10].
Lee, et al. [12] tested the FDM process as an aid in reducing tooling costs and
lead times associated with a typical IC process. High tooling costs for master
dies is not usually justifiable for small run production unless it has a special
application, e.g. defence industry, or is unable to be made by other
processes. A benchmark model consisting of common shapes and profiles
14
was created to assess the effectiveness of the FDM process for indirect IC
mould production with sacrificial RP patterns. The FDM process was shown to
be [12] effective for the economic mass production of complex metal parts
which may otherwise be difficult to produce.
RP patterns were also seen [12] as ideal substitutes, as they can be melted
and burnt out from the ceramic shell. Early RP wax patterns exhibited
cracking due to excessive thermal expansion when melted [11]. It was
discovered [12] that plastic ABS RP patterns chemically attacked the cavity
surface of the ceramic shell, due to corrosive degradation during pattern burn
out. This was overcome by simply creating patterns with a hollow internal
structure so that upon melting the pattern expanded inwards. Indirect ceramic
IC shells were then made by the FDM process, using ABS plastic on a
FDM3000 (Stratays Inc.) system. Both sacrificial FDM patterns and multiple
pattern production, using silicone rubber moulds poured over the FDM
pattern, were investigated. Thermogravimetric analysis for the residual ash
content was used to show how well the FDM patterns burnt out of the shell.
Testing conducted under O2 atmosphere at 900°C gave a residual ash content
of 2.218% for the FDM pattern compared to 0.04% for traditional foundry wax
material. However, Lee noted that actual burn out temperatures for the latter
was above 1000°C and, a lower residual ash content was to be expected.
Surface testing gave slightly lower values than that those presented earlier
[11]. The surface testing conducted on the cast Aluminium A356, gave 4.63-
4.69μm average roughness. Indirect mould production through silicon rubber
moulds was also analysed. The cast surface roughness was found to be
5.79µm, which was slightly higher than the sacrificial pattern method. In terms
of other defects, misruns were seen at 0.5mm wall thicknesses as well as
metal penetration. These two defects could be due to residual ash content but
misruns and metal penetration can be caused by incorrect filling and bad
formation of the ceramic shell respectively. Dimensional accuracy of the
sacrificial RP patterns was acceptable, with differences from actual dimension
to design dimension ranging from 0.75%-1.42% (dimensions ranged from 0.5-
110mm). Production costs were halved in both sacrificial and rubber mould
techniques. Overall, hard tooling was eliminated with the use of RP patterns,
15
which led to significant time and cost savings in small run production.
However, for large scale production, traditional methods were still preferred,
due to the ability to offset large initial tooling costs by the sheer volume of
parts produced.
Use of RP Moulds
Much of the early work on RP applications involved a variation in the IC
process by direct printing of thin ceramic shells. Elimination of sacrificial
patterns led to direct shell production of the ceramic shells and cores. Direct
Shell Production Method (DSPM). was pursued by Sachs, et al. [13]. Potential
applications for 3D printing were considered and moulds for metal casting
were assessed at an early stage as a high, potential production target.
Dimensional control and part strength were seen as the critical factors. Initial
investigation of 3D printed parts produced good dimensional control but low
part strength.
The traditional IC process produces high precision complex castings, but with
commercial disadvantages. These centred on high tooling costs to create new
dies, which produce the wax positives, and upon subsequent tooling changes
which increase lead times. The dies for wax positives can be made simply
from aluminium but those for abrasive ceramic cores must be made of harder
material such as carbide [13]. Figures of US$5,000-US$50,000 (1990) for die
sets were quoted as common [13]. This increases tooling costs, which are
further strongly linked to mould complexity. By using the 3D printing of
ceramic cores and shells, initial tooling costs for ceramic dies were eliminated,
making small to moderate production profitable. Sachs et al. cited lead time
as a critical factor in production.
Use of thin ceramic 3D printing moulds was developed by Curodeau, et al.
[14], to cast Cobalt-Chromium (CoCr) alloy to create the bony in-growth
surfaces for a medical application. The objective was to efficiently produce a
surface texture optimised for bone growth application. Orthopaedic
prostheses, e.g. knees and hips, require high precision metal parts which are
16
then implanted in the body. Standard production methods involved IC of the
complex pattern or pattern fabrication by five axis CNC machines. 3D printing
was incorporated to produce ceramic shells in hours without the need to
create a complex wax pattern. CoCr was then cast directly into the printed
alumina ceramic shell. Overall dimensional accuracy was found to be ±5-
10µm in all three build directions. The 3D printing process quickly created fast
complex metal parts, with intricate surface features. More significant research
was later undertaken to adapt the process for production of different textures
suitable for orthopaedic prostheses.
Notwithstanding developments in the 3D printing process [13], early attention
was more focused on the SLS process to produce moulds for metal casting.
By passing a laser over individual sand grains, localised high temperatures
result in the fusing of the material and binder to create moulds for casting.
Advantages were apparent [9] for design and design alterations using CAD
programmes. SLS processes reduced lead times and costs of design and
increased design complexity, with no significant increase in cost. The castings
produced from SLS moulds were found to be accurate and repeatable but
possessed poor surface finish and contained step patterns reflected from the
mould surface. Significant out gassing problems were present and further
testing and experience with the process was evidently needed to understand
these problems.
Direct production of sand moulds using SLS was also researched [10]. The
SLS SandForm process utilised Zircon sand and silica sand materials to
directly fabricate sand moulds from CAD files and eliminating the need for a
pattern. Laser sintering of moulds and cores was found to have equivalent
accuracy and properties to moulds and cores produced by traditional
methods. It was concluded that direct mould production methods such as SLS
reduced production lead times. Direct methods were shown to as increase
accuracy levels as intermediate replication stages such as pattern making
were eliminated.
17
Using SLS technology Gibbons [15] produced moulds for an intake manifold
for a KTM 525cc single cylinder engine. The moulds were produced from an
EOSint S700 SLS machine using silica Croning Sand (Phenolic resin coated).
Parts for this application were traditionally fabricated from sheet metal and
then welded. The traditional production process was expensive and suffered
from long lead times and limited design freedom. Total lead time for producing
the mould was 33 hours. The RP moulds were flame hardened on the surface
and then baked at 180°C to achieve full strength. The benefits of using RP for
sand moulds were: complete design freedom, true CAD to manufacture and
rapid design realisation. Also an adequate casting was produced for the
application and the mould had adequate permeability. Limitations reported
were poor surface roughness, which was assumed to be caused by a lack of
sand particle alignment and some geometrical limitations resulting from the
need to remove support structures from the mould.
Rooks (Rooks 2002), used the EOS S700 SLS machine at the Warwick
Manufacturing Group (WMG) and produced sand moulds and cores for three
V6 cylinder blocks. Casting was with aluminium and grey iron and compacted
graphite iron. The objective was to evaluate the cylinder heads for noise, heat
release and emissions and then compare findings to traditional prototypes. It
was found that the cylinder heads produced in the sintered sand moulds were
similar in all aspects to traditional prototypes but were completed in much
shorter times. WMG stated that this process reduced lead times from 4-6
months to 2-4 weeks and established a breakeven point of 100 RP castings
versus traditionally made prototype castings. They concluded that SLS for the
production of moulds was viable and ideal for complicated mould and cores
for small volume castings.
Tang, et al. [16] investigated the method of strengthening during SLS of silica
sand. This research focused on the solidification mechanism of the sintered
sand and the effects on accuracy, surface finish and strength, with scanning
speed and laser power as the primary variables. With the use of a Scanning
Electron Microscope (SEM), comparisons of un-sintered and sintered sand
grains were made to determine the strengthening mechanism of the mould
18
material. The binding mechanism was linked to the surface of the sand
particle, which was easily melted due to the presence of inclusions such as
Al2O3, which create a salt like eutectic with the SiO2. This reduced the melting
temperature of the sand allowing the sand particles to bind together through
the liquid surfaces and then solidify once the laser moved away. This process
was observed at different scanning speeds and laser powers. It was found
that the surface roughness and the compressive strength were proportional to
the laser power and inversely proportional to the scanning speed. The
sintered sand parts had good compressive strength and surface roughness,
with the accuracy of the moulds parts ranging from 0.1mm to 0.5mm.
However, the thermal nature of laser sintering created a heated zone which
caused shrinkage and distortion of sand grains, which was detrimental to the
accuracy of the mould dimensions.
Casalino, et al. [17] identified the influence of the main process parameters on
sand properties, permeability and compressive strength. These were
measured as primary responses with variables again being speed and laser
power settings. Experimental planning and analysis made use of a Taguchi
experimental design. LASER-CRON resin coated quartz sand particles (96.8
% quartz and 3.2 % resin) were cured at different times and temperatures in a
post heat treatment process on an EOSINT S 700RP SLS machine.
Compressive strengths were equal or better in terms of traditional mould
making processes and suitable for steel and aluminium casting with the limits
being stated to be between 50-150 KN/m2 for steel casting and 15-25 KN/m2
for aluminium casting. Permeability (represented in this case as a flow rate)
was also suitable for steel and aluminium casting, ranging from 160-190
cm3/min. This was considerably higher than referenced industry standards for
synthetic sand, which were 20-60 cm3/min for aluminium casting and 160-220
cm3/min for steel casting. The process variables of laser power settings and
post curing time were also influential in the fracture of sand specimens.
Rupture surfaces of sand specimens were observed to be angular, cone and
skin type. Shorter baking times produced skin type failures for lower strength
and cone type fractures for higher strength. Considering all of the variable
parameters, baking time was found to be the most significant variable on the
19
fracture type when evaluating the differences between skin and cone type
fractures.
Direct fabrication of moulds from CAD files for metal casting can be achieved
either by SLS or 3D printing. SLS sand sintering machines cost about
NZ$1,380,100 (based on EOS S700). The largest 3D printers cost around
NZ$120000. However, the SLS sand sintering machines are much larger and
can create larger shapes. Against that build times of 3D printers are quicker,
with the ability to cover entire areas rather than sintering of a single line laser.
Further, energy consumption is high with SLS technology due to the use of
high powered lasers. Even smaller SLS machines such as the EOSINT M 270
require 32 Amp‟s with a maximum 5kW of power needed. Large ZCorp 3D
printing machines consume around 1.8kW (based on the ZCorp 650).
As the 3D printing process increased in popularity, there was a renewed
research interest in establishing the advantages and feasibility of applying the
process to sacrificial mould production. Kochan [18] examined the cost
effectiveness of various RP processes, with particular focus on 3D printing. It
was found that limitations of 3D printing centred on the actual speed of
production itself, which was slow when compared to traditional moulding. The
size limitation of 3D printing machines to produce larger moulds was also
cited. It was noted there was an increasing use of RP systems, especially in
the automotive sector and especially with the use of low cost 3D printers. The
advantage of faster part production of ink jet 3D printing technology resulted
in some manufacturers (ProMetal) claiming larger build sizes and speeds ten
times that of common SLS machines.
Use of 3D printers for Part and Mould Production
Before the start of the 21st century, ZCorporation (ZCorp) 3D printing
machines had only offered plaster and starch based powders, which were not
seen as entirely suitable for metal casting. In the early part of the last decade
ZCorp developed a ceramic material which was produced specifically for
casting non-ferrous alloys. The ZCast „direct pour method‟ is an effective
20
method for production of aluminium and other non ferrous castings at high
speeds and low costs. The direct pour method produces pattern-less printed
moulds suitable for casting. Waurzyniak [19] noted that small production runs
of 10-20 parts can be achieved by using the ZCast process. The advantage of
the ZCorp process is that the prototype is created in the final material unlike
other RP systems which use plastic or sintered metals.
Research by Bak [20] showed the 3D printing technique to be a superior
process to other RP processes for producing components by single line SLS
technology. It was shown that 3D printing method allowed users to produce
tooling locally, either on site or at a local foundry to facilitate short run
production. Tolerance testing and surface testing of the printed moulds were
reported to be ± 0.38mm and 200-300µm respectively (100µm when a mould
wash was used). The use of the ZCast process was found to need at least a
3mm mould wall thickness in direct pour casting applications. Additional
information [20] on run size was also presented, and production of a
dispensing manifold was achieved with fewer than 50 production units needed
to break even. It was thus cheaper to purchase a 3D printer and use the
ZCast process rather than use traditional methods such as creating a pattern
and then creating a mould. The use of 3D printing technologies was seen to
be increasing sharply, perhaps at the expense of SLS and other RP
processes associated with the casting process.
Rebros, et al. [21] undertook research (on a rival 3D printer, ProMetal) which
focused on the thermo-mechanical properties of 3D printed moulds. Due to
the increasing focus on production processes to produce near net shaped
parts, with thin wall and tight dimensional controls, he investigated distortions
in chemically bonded 3D printed sand. Thermal Distortion Testing (TDT) was
used to evaluate changes produced by the thermo-mechanical reactions of
the chemically bonded sand. It was shown that when molten metal comes in
contact with shaped sand the heat transferred from the hot metal causes
thermo-mechanical reactions, which result in dimensional changes to the
sand. The TDT was conducted on 3D printed Silica sand using a furan binder,
printed on a ProMetal S15 rapid casting machine. For comparison, chemically
21
bonded silica sand using the Phenolic-Urethane Cold box (PUCB) process
was studied. TDT showed thermo-mechanical and thermo-chemical changes
at elevated temperature (760°C) for both sand systems. Mass loss and
surface cracking were observed in all 3D printed sand and PUCB material
specimens. In all cases it was found that the specimens showed signs of
cracks and elevated pressures and temperatures resulting in distortions of
both 3D and PUCB sand systems. The cracking witnessed in all specimens
was explained by expansion and contraction differentials in the sand
composite. Overall, the testing showed that after 90 seconds, there was no
significant difference in the two sand systems regarding mass loss (%).
However, for shorter high temperature exposure the 3D printed sand lost less
mass than the PUCB sand. More importantly, larger total magnitudes of
thermo-mechanical changes were seen in the PUCB sand in the 3D printed
sand. This was attributed to a higher level of thermo-chemical reactions in the
PUCB sand system.
More recent primary research on the 3D printing process was mainly been
concerned with the mould accuracy and mould properties for casting. Bassoli,
et al. [22] researched the feasibility and accuracy of the 3D printing process
with two similar techniques. These were included a printed sacrificial ZCorp
starch pattern for further use in IC and a directly printed ZCast mould. An
automotive part was chosen for casting and subsequent dimensional analysis
was carried out on a Coordinate Measuring Machine (CMM). Printed starch
patterns used in IC were infiltrated with wax and once removed Al-Si
(aluminium–silicone) 304 steel was cast. Printed ZCast501 moulds were used
to process Al-Si10% alloys. The printed moulds were heat treated at 6 hours
and 200°C. Surface roughness measurements were carried out on both the
internal and external surfaces of the cast part. The casting obtained from the
invested starch pattern exhibited good surface quality with average roughness
values (Ra) of 4µm. Surface porosity and filling of the mould was said to be
acceptable. However, the cast part produced in the printed moulds showed
both signs of micro-porosity and a parting line causing an edge on the outer
surface. The average surface roughness was found to be 10μm. Results from
the overall accuracies of both the starch pattern and cast parts were similar.
22
External International Tolerance grades were IT15 for the starch pattern, IT16
for the IC and IT15 for the ZCorp Al casting. It was concluded that both
methods were effective in that castings could be produced quickly and
adequately. The ZCast process provided satisfactory results but was limited to
the range of light alloys. It was found that geometrical freedom was increased
in the ZCast process, with the surface roughness noted as the only limitation.
More analysis followed [23-25], researching the capability of the 3D printing
process to create wax patterns for further use in IC. Difficulties in creating
complex parts were experienced, such as the warping of starch patterns.
However the alternative CNC machining of a wax mould cavity was seen as
more difficult, expensive and time consuming. The relative strengths of 3D
printing process were found to be high speed, ability to print in colour,
geometric freedom, and large build sizes (3D System printers).
Disadvantages were inaccuracy when compared to other RP systems, poor
surface finish of the mould material, limited material options and secondary
epoxy and baking processes.
Further in depth analysis of the dimensional accuracy of the printed ZCorp
parts was conducted with a CMM on different RP materials [25]. Various
distances were evaluated and comparisons were made in the x,y and z axes
of the same part. Starch (ZP14) and plaster (ZP100) materials were trialled on
a Z400 3D printer. In order to quantify the accuracy of the printed parts a
benchmark model was created to assess accuracy indicators. A second part,
a differential housing, was also printed. Accuracy indicators included surface
profile, circularity, concentricity and angular tolerance. Results showed that
primary factors responsible for deviation in measured accuracy were material
type, build axis and magnitude of measurement. Measurement tolerance of
this 3D printer was IT9-IT16. Plaster based powders were found to yield
higher accuracy. The probable reason for this is the finer grain size allowing
for finer layers of 0.1mm over that of 0.18mm for the starch material. With
both materials parts were found to be printed slightly larger than the CAD
models but skilled selection of scaling was reported to reduce this problem.
Further research by Dimitrov, et al. [24] looked into improving design and
23
manufacturability of foundry equipment by using 3D printing to produce
moulds and cores for casting. An evaluation of directly 3D printed shell
moulds with ZCast501 and printed sand cores for use in traditional moulds
was undertaken. Ten moulds were produced with a mixture of traditional
foundry methods, using ZCast cores and ten directly printed ZCast501
ceramic shells. Surface roughness testing was made on ten different surfaces
for each casting. Results showed that surface roughness (Ra) was 13.6μm for
traditionally made castings and 14.9μm for castings produced in ZCast
ceramic shells. The layered production technique led to steps on the printed
mould surface, which was reflected in the castings. However, surface
roughness tolerance was found to be acceptable for sand cast parts. Overall,
it was found the use of 3D printing allowed greater flexibility than traditional
foundry mould production by allowing quick manufacture of moulds and cores.
Mould properties such as permeability and compressive strength of the ZCorp
ceramic material (ZCast501) were researched by the current author, [26] as
part of undergraduate project work. Also, the suitability of mould coatings and
their effects on surface finish of the castings were investigated. Basic
mechanical data was first gathered to compare to traditional foundry data.
Initial testing was conducted on grain size distribution. It was stated [27] that
common foundry sands have average grain sizes of around 220-250µm,
whereas the average grain size distribution of the ZCast501 material was
about 100µm. The latter has a detrimental effect on the surface quality of the
castings. Due to the large range of the grain size 50-300μm there was metal
penetration into the moulds giving a poor surface finish to the cast specimens.
The effect of mould baking on ZCast501permeability and compressive
strength was examined. Time and temperature was varied from 3.33-8.8
hours and 130-275°C for the investigations, based on a two factor central
composite experimental design. Optimum permeability and compressive
testing of the mould samples gave 8cm2/s and 1MPa respectively.
Permeability was low and this was attributed to the large grain size
distribution. Compressive strength was found to be acceptable. Further
analysis by SEM of the fracture surfaces from compressive testing, revealed
decomposition of the bonding strength of the gypsum plaster to the larger
24
sand grains. The smaller gypsum particles acted as a solid binder much like
clay in traditional green sand moulding. It was suspected that additional time
and temperature curing removed excessive water, embrittling the gypsum and
liquid binder. This sharp reversal in trend with increased heat energy was
initially thought to be due to some of the material melting, fusing or changing
phase. This would mean that the intergranular pathways would become
blocked or at least partially blocked, thus decreasing permeability. From the
SEM work conducted, specimens baked at higher temperatures showed a
significant change in terms of the gypsum structure and morphology.
Bassoli and Atzeni [28] experimented with direct metal casting into ZCast
moulds to optimise the mechanical properties of the cast products and
calculate suitable tolerance grades of the cast parts. Cylindrical specimens
were printed out and cured with varying times and temperatures. The baked
parts then underwent compression testing and the rupture surfaces were
evaluated with SEM. Time and temperature varied from 160-250°C and 4-8
hours respectively. A benchmark pattern was produced for further evaluation
of dimensional tolerances. Upon baking, the printed part colour was found to
change from white to dark brown. Changes in dimensions resulted from the
times and temperatures of baking showed small insignificant changes. The
green parts (before baking) were seen to be the most inaccurate. In
compressive strength testing the green parts were seen to have the highest
strength but strength decreased as time and temperature of baking increased.
Compressive strengths of the baked samples ranged from 2.6-6.2MPa, which
is a much higher order of magnitude than the current author found from his
own testing (around 1MPa). Thermogravimetric analysis showed temperature
as the primary influencing factor for compressive strength. Samples were
seen to lose 6-7% weight after baking at 4hours at 150°C. The current author
however found the opposite, with exposure time more significant than baking
temperature. Differences in testing and experimental design may account for
some differences along with natural variation. An international tolerance (IT)
grade of 15 was assigned, which was consistent with foundry applications.
This tolerance grade was also in line with other findings [23] who quoted IT9-
IT16. After completing accuracy tests it was concluded that the moulds were
25
described as „fundamentally the same‟ in terms of dimensional variation both
before and after baking. Dimensional accuracy was said to be the same in all
three build directions with all tolerances calculated with 95% confidence.
In casting Zinc (Zn) alloy in printed ZCast501 shell moulds, Kaplas and Singh
[29] found similar tolerance grades to that of Al (IT13-15). Tolerances were
found to be acceptable and shell thickness could be reduced from 12mm to
just 2mm. Radiographic testing of the castings showed decreased shrinkage
and gas levels in the casting at lower shell wall thicknesses. Cost and times
savings of 41 and 37 percent respectively were observed when using the
2mm wall thickness over the recommended 12mm. In a further report, Gill and
Kaplas [30] compared printed shell moulds using the ZCast501 powder and
the 3D printing technique for the creation of sacrificial starch and plaster
patterns to be used in the IC process. Moulds and sacrificial patterns were
printed on the ZCorp 310 Plus 3D printer. Sacrificial patterns once printed,
were infiltrated with liquid acrylate material. Al was cast and finished casting
dimensions were examined for accuracy by a CMM. The thickness of the
printed shells was also varied from the recommended 12mm down to 2mm.
Accuracy of the cast parts produced by IC and printed moulds were similar.
The investment cast sacrificial patterns had an IT of 14, with printed shells
ranging from IT13-16. This was similar to previously reported tolerance
ratings, [25, 28] indicating that this tolerance rating is appropriate for ZCorp
3D printers. Results from surface testing showed that the surface roughness
(Ra) values of sacrificial patterns ranged from 3.99-4.21μm and that of
samples from the printed shells ranged from 6.78-6.98μm. These figures are
in line with other surface roughness values presented [11, 12] when using
sacrificial patterns. Results from reducing shell thickness showed that
reducing from 12mm thickness to 6mm gives better dimensional and
mechanical properties (with lower thicknesses showing more variability in
dimensional accuracy). Likely reasons for this were increased rate of heat
transfer due to reduced wall thickness of the shell.
In similar research, Singh and Singh [31] looked into the accuracy of the
printed ZCast shell mould for casting lead. A thrust pad for a two stroke petrol
26
engine was used as the pattern to be cast. Shells were built on a ZCorp 510
printer and cured at 110°C for one hour. Tolerance grades varied from IT10-
13 with an optimum shell wall thickness of 1mm. This reduction in wall
thickness leads to decreased costs (material reduction) of around 45.75% and
a 43% decrease in production time when compared to the 12mm shell
thickness. Evidence of improved dimensional accuracy and increased
mechanical properties was shown by CMM measurements and
photomicrographs respectively. The refined grain structure was most likely
due to faster solidification resulting from decreased shell thickness, and the
consequent higher heat transfer rate.
From the examined literature it is clear that the use of three dimensional
printing of moulds and cores has been researched mainly to investigate the
accuracy and properties of printed moulds. Mechanical properties, accuracy
and quality of the castings have been looked at to a lesser extent. However
the data on the casting quality is primarily limited to aluminium prototypes.
Magnesium Casting
Due to the presence of gypsum in the moulding material, which has a low
melting temperature, there has been no research into applying 3D printed
moulds to ferrous metals and alloys. Further, there is also little evidence of
research into other non-ferrous metals. In particular, there has been renewed
interest in Magnesium (Mg) and in recent years. Global warming and
sustainability issues have meant that engineers and scientists incentive have
to create more efficient, sustainable and recyclable processes to avoid
excessive energy consumption and pollution. One of the largest contributing
sectors to overall world CO2 levels is the transport sector. Burning of fossil
fuels not only produces CO2 which contributes to the greenhouse effect, but
also other harmful emissions. These include Carbon Monoxide (CO), Oxides
of Nitrogen (NOx), Hydrocarbons (HC) and Oxides of Sulphur (SOx). Several
directions of research to reduce these emissions have lead to the
development of bio-fuels from waste, hybrid engine technology and hydrogen
fuel cells. One of the major targets in current car production is reduction of
27
mass. Small reductions of 5-10 % for each vehicle would result in very large
reductions of CO2 if implemented worldwide.
Since the early 1960‟s the use of Al in automotive applications has slowly
replaced steel due to large weight savings and increased performance. Mg is
now being assed as a possible replacement for Al, with increasing research
into the mechanical properties and limitations of Mg alloys, specifically in the
context of processing by sand casting. The International Magnesium
Association (IMA) reports that use of die casting Mg in automotive industry is
increasing at an unprecedented annual rate. Eliezer, et al. [32] underlines this
claim, explaining increased demand has arisen from the automotive sector.
Further, large supplies of Mg exist with an estimated 22000 tonnes (t) year
supply in the Dead Sea alone available based on current production rates
[32]. A periodical article by Eigenfeld, et al. [33], highlights the known
advantages of Mg, particularly weight reduction. He reported that 40000t of
metal was used in die casting alone in 1995 and by 2001 this figure had
increased to 125000t.
The potential of Mg to reduce weight and increase efficiency has lead
research into the applications and advantages of Mg alloys.
Bronfin and Aghion [34] presented several reasons to use Mg rather than
other metals. Mg alloys are the lightest structural metallic material, and
therefore of attraction for automotive and aerospace applications. With
increasing emphasis on reducing emissions and sustainable transport, many
car manufacturers are moving to use 40-100kg of Mg alloy in each vehicle
[34]. This means significant weight savings when compared to traditional steel
engine blocks and housings and to a lesser extent Al. Reduction of weight in
turn reduces fuel and energy consumption. In one example [34], the block,
wheels, front cradle and gear housing were constructed in Mg which led to a
fuel saving of 0.25 litres/100km when substituting steel and 0.1 litres/100km
when substituting Al. Additionally, Mg alloys have good electromagnetic
interference shielding properties making it an attractive material in audio and
electronic equipment. High strength to weight ratios and thin wall casting
capabilities also lend Mg of use to the electronics industry. Further
28
characteristics [34] are good heat dissipation, dimensional stability, Radio
Frequency Interference (RFI) shielding capabilities, damping capabilities and
recycling capacity.
Mg also has very good machinability due to its Hexagonal closed package
(HCP) structure with a limited number of slip planes. The HCP crystal
structure and favourable atomic size allows Mg to form a solid solution with
various elements especially commercial elements such as Al, Zn, Li, Ce, Ag,
Zr and Th. „Workhorse‟ alloys such as those based on the Mg-AL-Zn system
alloys such as AZ91 have excellent strength and acceptable short term
elevated temperature properties. It has been noted [34] that creep strength
and bolt load retention were poor properties of Mg-Al-Zn alloys but these
issues were then overcome by Volkswagen AG and Audi AG for die cast gear
box housing by design modification. Inadequate creep properties of Mg have
also led to use of Rare Earth (RE) alloys to allow for higher temperature
applications. Mg-Al–Si alloys are high creep resistant alloys (e.g. AS21 and
AS41) and potentially able to provide good castings. Low Al content has led to
casting difficulty and poor corrosion resistance. To ensure good fluidity Mg
must have a significant amount of Al but Al also leads to the formation of the
eutectic Mg17Al12 intermetallic, which has lower hardness and an adverse
affect on creep properties. Corrosion is another issue for Mg use, especially
problems with galvanic corrosion but good design is cited as a means to
overcome this. Generally Mg alloys such as AZ91 have good corrosion in
atmospheric conditions but chloride environments are a concern to the wider
application of Mg [32].
Casting is the primary manufacturing process for producing Mg in particular,
die casting which offers many advantages. These advantages include a high
fluidity (good castability), and low specific heat which reduces die wear. Iron
also has low solubility in Mg which reduces sticking to metal dies. Higher in-
gate pressures can also be achieved at lower overall pressures due to low
density of Mg. However some recent research with sacrificial moulds such as
sand (green and chemically bonded), plaster and IC shells has been
undertaken. Much of this work has focused on how to increase the quality of
29
the cast parts. An important component of Mg casting in a sacrificial mould is
mould-metal reaction. Severe oxidation of Mg in air has lead to research of
inhibitors and coatings for refractory mould materials to reduce mould metal
reaction. Inhibitors such as Teflon (C2F4), sulphur (S) and boric acid (H3BO3)
and potassium borofluoride (KBF4) have been successfully tried by Hanawalt
and Okada [35] in low water and oil (2-3%) bonded green sand when casting
AZ91C. Further work [33] saw the use of sulphur acid, boric acid, potassium
tetrafluorborate and ammonium fluorosilicate used as inhibitors to prevent
reactions. Green sand mixtures of 6% bentonite clay and 1.5-2% moisture
content and 3% inhibitor were found to be optimum ratios [33]. Further,
optimum compactability of green sand moulds was 40% and increasing water
content was found to strongly increase mould metal reactions. It was also
shown [33] that for chemically bonded furan/acid nobake sand system
comprised of silica sand, mould metal reaction was inhibited by the spaying of
an alcohol based coating containing iron oxide.
Fantetti, et al. [36] working with Mg and plaster moulds showed that for
adequate results, moulds have to be thoroughly dried and moisture free. Lun
Sin, et al. [37] showed that when molten Mg enters the mould, any water
present will result in steam which will produce hydrogen and ignite in oxygen.
Sufficient baking times and temperatures are outlined in order to minimise
free water content and a minimum temperature of 105°C was recommended
for plaster moulds. It was found however, that increasing the mould pre-heat
temperature led to more severe mould-metal reaction. This was attributed to
the decrease in the temperature gradient between the molten metal and
mould material. This therefore increased the time period at which the molten
metal was exposed to higher temperatures, promoting reactions to occur,
similar to those found by Idris and Clegg [38].
Expanding further on mould metal reaction, a recent report by Takamori, et al.
[39] looked at the reaction of Mg with silica green moulds. When Boron Nitride
was used as a mould wash on the silica mould, no reaction was seen. The
effect of pouring temperature was also found to have an effect on the cast
surface with higher temperatures promoting blacker and rougher cast surface
30
finishes. Observed reactions were however most likely due to the moisture
present in the moulds when the moulds were cooled down to room
temperature [38].
The study of mould metal reaction has also been reviewed with respect to IC
Mg. The use of more thermodynamically stable oxides such as Mg oxide
(MgO), Calcium oxide (CaO), Calcium carbonate (CaCO3), Aluminium oxide
(Al2O3), Zirconium Orthosilicate (ZrSiO4) and Calcium Zirconate (CaZrO3)
have been experimentally trialled [37, 40, 41]. Lun Sin et al. [41] however
found that grading of oxides by free energy change was not entirely accurate
and that mould pre-heat temperature was again a critical variable in the
reactions that occur at the mould metal interface. Testing on stable oxides by
Cingi, et al. [40] showed large differences in free energy change between the
reactions involving no free oxygen and with free oxygen. The free energy
change is over 3.5 larger when free oxygen is available. It was concluded [40]
that oxygen is not just necessary for initiation of mould metal reactions but
also for their continuation and formation of other products of reaction.
The increasing use of light metals, led by Al, highlights their ability to reduce
weight and energy consumption whilst still satisfying mechanical demands.
The unique ability of light alloys to provide high strength to weight ratios has
seen increasing use in many engineering sectors notably the transport and
aerospace industry. In addition, light metal applications worldwide are also
vital for reducing emissions and achieving a more sustainable society. There
has been an increase in research towards improving the processability of light
metals. The foregoing literature looked at the use of rapidly produced moulds,
clearly highlighting the improvements in design freedom, time and cost
savings in small and medium volume production and in particular, paving way
for rapid casting. A number of studies have identified that the tolerance limits
attainable by the 3D printing process were within the normal limits generally
obtained in traditional foundry processes. The relative advantages of using
the 3D printing process over SLS in terms of speed and cost, have led to an
increased use of 3D printing in mould and core production. Results from the
preliminary analysis of the ZCorp ZCast501 material, showed that grain shape
31
and distribution led to slightly higher cast surface roughness and reduced
permeability. However, with the application of mould coatings the surface
roughness was improved to adequate roughness values. Further, mould
compressive strengths were found to be adequate and tensile properties of
the castings were acceptable when compared to traditional sand casting.
Overall, the experimental results showed that a proper combination of process
parameters and use of a suitable coating material will bring the mould
characteristics close to the values recommended for traditional foundry
practices for light metals.
Once the mould material characteristics are identified and standardised, it
becomes important to experiment, analyse and understand how different
metals would react with these moulds when cast. Light metals like Al and Mg
are of particular interest due to their advantages noted above. While there is
evidence of Al being researched with some of these techniques, other light
metals have attracted very little or no attention. One probable reason for this
is that quite a few of these RP techniques are relatively new. Renewed
research interest in metals like Mg processed by sand casting is a relatively
new development and there are obvious gaps in the literature. Interesting
research questions arise as to what happens if metals like Al and Mg are
processed to produce castings using 3D printed moulds. What are the mould
metal reactions? What are the mechanical characteristics of the cast parts?
What is the optimum combination of process parameters for the best
performance of the rapid casting process in general?
Pressure Filling Techniques
Better mechanical properties can be achieved for light metal castings by using
a pressure filling, as in the case of high pressure die casting. This produces
thin complex sections which have superior mechanical properties. Gjestland
and Westengen [42] showed that grain size and Secondary Dendrite Arm
Spacing (SDAS) are strongly related to the solidification rate in Mg and less
so in Al, which is usually high in die casting due to the rapid cooling effect of
the metal dies. Fast solidification from the chilling die materials yield a fine
32
grain structure and improved SDAS giving the optimum mechanical
properties. More importantly, the application of high pressure allows filling of
complex and thin sections. Tensile yield strength of Mg follows a Hall-Petch
relationship, meaning that with decreasing grain size, grain boundaries are
strengthened through reduction of dislocation movements of individual grains
[42]. The characteristics of castings produced by HPDC are thus improved
ductility and increased tensile properties with a refined grain structure. Also,
solidification shrinkage is reduced due to the high pressure, which facilitates
feeding during solidification.
Gravity casting processes such as IC and sand casting cannot attain the
pressures created in HPDC. However, Centrifugal Casting (CC) offers a
possible solution, to bridge the gap between gravity filling of a traditional sand
casting process and the pressure filling of the die casting process. This results
in more sound and dense castings. Since the metal experiences pressure,
conventional casting defects such as internal shrinkage, gas porosity and non
metallic inclusions are less likely to occur in CC. Improved mould filling in CC
also leads to a reduction of rises and runners, reducing scrap rate and
secondary fettling processes. Also natural directional solidification occurs as
the solidifications starts from the outside and moves inwards to the central in-
gate. Sacrificial moulds and cores can be also used in the CC process with
green or dry sand, plaster and graphite moulds. This opens up a new avenue
for experimentation with RP moulds. Combining 3D printing as a mould
making process and CC as a filling technique may produce advantages from
pressure filling, paving way for sound RP castings. Current research is
therefore aimed at the feasibility of CC in rapid prototyped moulds. A brief
review of the CC process and related matters follows next.
Specific variables in the CC process include melting practices, mould and
pouring temperatures, rotation speed and alloys. Most importantly, rotational
speeds can be varied greatly. With respect to green sand moulding, low
speeds are recommended for initial pouring with gradual increases up to the
rated speed after partial filling of the mould. This is combined with the
application of a mould coating or wash to green sand mould to ensure a dry
33
mould surface. Strieter and Maonner [43] looked into CC of light metals like
AZ92 Mg alloy into steel dies using a vertical CC process as early as 1946.
The use of Mg alloys in CC was very limited, due to rapid oxidisation upon
melting and severe mould metal reactions. Furthermore, density of molten Mg
is lower than that of Mg oxide, which means that oxide entrapment is common
and difficult to remove. Test results showed several problems including pits,
micro-porosity and cold shutting. Pitting was found to be due to excessive
turbulence of the molten metal. In an attempt to reduce this effect, metal was
initially poured at low rotational speed and then after filling with increased
speed. No significant improvements were seen. The creation of oxides and
other defects were also said to be due to excessive turbulence, especially at
increased temperatures. Despite these defects, optimum process variables
were found to be around 1000rpm and 700-760°C. Mould temperature was
315°C. Cost and mechanical properties of the castings were of equal or better
quality than static permanent mould castings. Ultimate Tensile Strength (UTS)
properties ranged from 252-255MPa compared to that of 206-215MPa for the
equivalent static casting. Grain size was found to vary from 76.2-114μm when
compared 228-279μm for the equivalent static casting. Ductility was found to
vary from 1-1.8% compared to 1-1.3% for the static casting. All mechanical
properties tested showed improved properties with the CC method.
Mechanical property improvement in CC was also observed by Chirita [44] in
relation comparison to a static gravity casting model. During cooling, the first
nucleation sites for solidification in CC come back into the melt, thus
increasing the number of solidification sites due to excessive movement of
lower temperature particles. This leads to quicker solidification and a refined
grain structure when compared to gravity casting. This was similar to vibration
of a static mould during solidification. Photomicrographic evidence [45]
suggested that increasing rotational speed significantly refined the coarse α-
aluminium grains to fine equiaxed α-Al grains. Higher rotational speeds also
meant uniform distribution of secondary phases, leading to improved
mechanical properties and increased wear resistance.
Mathematical modelling of the whole centrifugal process is important as this
provides insight into the mechanics of the process where pure experimental
34
techniques become more difficult to test. Most recent work is centred on the
application of numerical models to analyse fluid flow and thermal fields in CC.
While most analytical work was on the mechanics of functional grading of
materials using CC, Howson [46] analysed horizontal and vertical centrifugal
casting to considering the fluid dynamics of the process. His model
considered forces acting on a liquid particle in contact with the mould wall and
excessive turbulence and particle separation were shown as possible defects
in CC. In 2007, Mesquita et al [47] developed an analytical model considering
an infinitesimal element of liquid under the action of centrifugal and friction
forces, after experiencing an initial velocity due to gravity filling through the
down gate. A governing differential equation was developed taking into
account the friction force and the centrifugal force as functions of time, to
determine the radial position and velocity of the molten metal from the central
down gate at different time intervals.
Summary of Literature
From the above literature, the application of RP in the casting industry has led
to significant research, primarily into RP pattern use and its suitability in the IC
process. This research related to pattern accuracy and the ability of patterns
to be invested and burnt out. Later use of RP techniques utilised sacrificial
moulds produced by the SLS and 3D printing processes and this prompted
investigation into achievable tolerances of printed moulds. Some work was
also conducted on the cast part quality, with studies focusing mainly on the
surface quality and cast part accuracy. Little work was conducted on the
actual mechanical properties of the castings produced in 3D printed moulds
with Al alloys and the author found no literature available on casting other light
metals such as Mg. Further, the suitability of RP moulds in pressure filling
applications such as centrifugal casting is unknown, with little literature
quantifying the effects of process variables when using the this technique.
35
Project Objectives Identified from Literature
Initial investigations looked to quantify the effect of varying temperatures and
time baking factors on ZCorp material (ZP131) to establish optimum mould
compressive strength and permeability. These values were then compared to
work previously completed on ZCast501 material.
Further investigation is aimed at metal filling under gravity (static) pressure to
test the impact on subsequent cast mechanical properties.
Investigation into Centrifugal Casting (CC) follows, to show that increasing
pressure filling of liquid metal is practically feasible with these moulds. This
will then permit combining the advantages of 3D printing technology with the
benefits of pressure filling.
The research then progresses to the actual response of RP moulds to both
static and centrifugal pressures, mould erosion, metal-mould reactions and
gas porosity. The significance of static and CC parameters on the mechanical
properties will be examined using these RP moulds and processes.
This work will thus identify the best combination of mould material, alloy type,
mould coating and pouring temperatures to produce acceptable castings.
These optimum parameters will then be used for processing centrifugally cast
parts in RP moulds. Influencing factors such as rotational speed, cavity angle
and in gate ratio will be investigated. A statistical design of experiments will be
employed to facilitate conclusions on the effects of multi-process variables
upon the mechanical properties of castings produced.
1.4 Research Questions, Hypotheses and Project Objectives
Broadly, the current research comprises experimental testing to establish the
properties of 3D printed moulds and light metal castings when used in the
static and CC processes. Specifically, the proposed research questions are:
36
(i) What happens to the moulds and metals utilised like Al and Mg when
static and centrifugal processes produce castings using 3D printed
moulds. Specifically, what are the mechanical characteristics of cast parts
and the optimum combination of process parameters for the best
performance of the cast parts?
(ii) What would be the response of these moulds and castings to increased
centrifugal pressure? More specifically, what effect does rotational speed,
cavity orientation and runner-cavity ratio have on the mechanical
properties of the castings, and how do these castings compare to
traditional static castings?
The following are the proposed hypotheses:
(i) It is postulated that each 3D printed mould material has an optimum set of
process parameters for the best combination of mould characteristics.
(ii) It is also thought that an optimum set of process variables exist for each
casting process, from which the best mechanical properties will result.
Specific Project Aims
Establishing the critical properties of 3D printed moulds
Establishing the suitability of these moulds to process light metals, such
as Al and Mg and the differences that exist when compared to traditional
sand castings
Establishing the mechanical properties of the castings produced in these
moulds
Determining the key influences in both static and centrifugal casting
process and quantifying their effect on the mechanical properties of the
castings
37
1.5 Methodology
Much of the work conducted throughout this project consisted of experimental
trials and interpreting their significance within a statistical model. The
objective was to produce quantitative information and the optimum
combination of process factors to be used for further experimental work. This
methodology is divided into three sections covering the experimental testing
conducted on the mould materials and the work involved in the static and CC
trials.
The first methodology was to analyse the suitability of 3D printed moulds for
producing metal castings by introducing time and temperature baking
variables. Tests were conducted on the moulds to evaluate mould
permeability and strength of the rapid prototyped materials, namely ZP131
plaster material and ZCast501 material (previous work). These experiments
were conducted in the context of a central composite experimental design
which was able to fit an experimental model to determine the significance of
different factors and establish an optimum combination of these factors for the
mould responses.
After these initial mould properties were established and the optimum
combination found, static casting trials involving two magnesium alloys and
one aluminium alloy were cast in these mould materials and also in a
traditional silica foundry sand material. A basic mould design consisting of a
central inlet, pattern and riser was created, in which the pattern consisted of
cylindrical specimens which were machined down to the appropriate tensile
test sample. This mould design and the castings produced were then
analysed using a Taguchi L9 experimental technique. These static casting
trials involved varying a number of process factors to establish their influence
on the mechanical properties of the cast metals, and the suitability of these
moulds to process these castings. The factors included the pouring
temperature of the alloys, the mould coating, the mould material and the alloy
type. Signal to noise ratio analysis was then undertaken using the Taguchi
technique, which resulted in the ranking of factors and their influence on the
38
mechanical and microstructural responses. A subsequent Analysis of
Variance (ANOVA) was then undertaken to establish the level of confidence
for the significance of these factors. Finally, the most influential process
factors were correlated, so that further casting trials could utilise these
settings.
Following the static casting trials, the suitability of these moulds in the
centrifugal casting process was examined. The ability to create increased
filling pressures in the centrifugal process was used to determine its influence
upon the mechanical properties of the cast parts. From the previous work on
static casting trials, the optimum mould material and process factors were
used in the centrifugal castings. The lack of experience in this process led to
some initial trials being conducted at the supporting company, Centracast.
After these initial castings were completed, and suitable centrifugal process
variables identified, an experimental was designed in which process factors
such as the speed of rotation and geometry mould cavity were evaluated. This
was to determine their influence on the as-cast mechanical and
microstructural properties of test specimens. The completion of the
experimental analysis and ANOVA on the resulting model established the
relationship of process factors to the mechanical properties of the castings in
the centrifugal process.
39
CHAPTER 2 CHARACTERISTICS OF 3D PRINTED MOULDS
2.1 Mould Materials
The mould materials supplied by the ZCorporation (ZCorp) are proprietary
materials and little information is available regarding their essential
characteristics for metal casting. The critical properties of moulds are:
Mould Compressive Strength (MCS)
Mould Permeability (MP)
In this study less important characteristics such as refractoriness and loss of
ignition are not analysed. MCS and MP need to be optimised in order to
ensure that adequate castings are produced. Optimisation of both variables is
obtained by determining the baking times and temperatures of the mould for
the highest MCS and MP. Thus experimentation proceeds by way of
quantifying the mould behaviour with respect to time and temperature of
baking so as to develop models to predict MCS and MP for use in future
mould design. The mould materials used are plaster (ZP131) and a specially
formulated ceramic, (ZCast501), with the latter being recommended by the
manufacturer for light non-ferrous castings. These materials are chemically
bound by the relevant binder, being ZB60 for ZP131 and ZB58 for ZCast501.
These are baked in a furnace to achieve its dry strength1.
2.1.1 Basic Ingredients of the Mould Materials
Basic ingredients of the ZCorp mould materials and binders used have to be
established either by independent testing or from the data specified in
Materials Safety Data Sheets (MSDS).
1 Dry strength refers to mould strength obtained after subsequent baking and curing
40
Mould Materials
Independent testing of ZP131 by X-ray diffraction revealed the following
ingredients as the basic constituents of this material:
CaSO4·0.67H2O
This composition is partially Calcium Sulphate Hemihydrate [48].
Figure 2.1 Spectra of ZP131 mould material by X-ray diffraction
XRD testing of ZCast501 also revealed the following ingredients as the basic
constituents of this material:
1. CaSO4·0.67H2O (50 wt. %)
2. MgSiO4 (50 wt. %)
The first composition is partially dehydrated Calcium Sulphate, known as
Calcium Sulphate Hemihydrate [48], commonly referred to as gypsum plaster
or Plaster of Paris. The second composition is Olivine Sand.
74-1639 (C) - Calcium Sulfate - CaSO4 - Y: 35.25 % - d x by: 1. - WL: 1.5406 - Orthorhombic - a 6.23000 - b 6.98000 - c 6.97000 - alpha 90.000 - beta 90.000 - gamma 90.000 - Base-centred - Bbmm (63) - 4 - 303.
Operations: Background 1.000,1.000 | Import
d:\guests\AUT-plasterAl.RAW - File: AUT-plasterAl.RAW - Type: 2Th/Th locked - Start: 15.000 ° - End: 70.000 ° - Step: 0.020 ° - Step time: 1. s - Temp.: 25 °C (Room) - Time Started: 0 s - 2-Theta: 15.000 ° - Theta:
Lin
(C
ou
nts
)
0
100
200
300
400
500
600
700
800
900
1000
1100
1200
2-Theta - Scale
15 20 30 40 50 60 70
41
Figure 2.2 The XRD spectra graph showing the refraction angles of the ZCast 501
material
Comparative Manufacturer MSDS of Liquid Binders
To bind the 3D printer mould materials together, a proprietary binder
produced by ZCorp is applied as an adhesive between each material grain.
Analysis of binders is not amenable to powder X-ray diffraction but MSDS
gives a good indication of the binder composition. Table 2.1 and Table 2.2
show the basic ingredients and compositions of the binders used for ZP131
ZCasy501 respectively.
While the basic chemistry and bonding mechanisms in each of these material-
binder systems are important, there is little existing literature on the
effectiveness of the ingredients and the proportionalities of the materials used.
In this research, however, these aspects are not considered, and attention is
mainly focussed on the overall characteristics and response of the mould
material aggregates under varying process conditions for metal casting.
d:\guests\StG-sand.RAW
85-0531 (C) - Calcium Sulfate Hydrate - CaSO4(H2O)0.67 - Y: 68.75 % - d x by: 1. - WL: 1.5406 - Monoclinic - a 12.02800 - b 12.67400 - c 6.92700 - alpha 90.000 - beta 90.000 - gamma 90.210 - Base-centred - B2 (5
80-0944 (C) - Olivine - Mg2(SiO4) - Y: 56.25 % - d x by: 1. - WL: 1.5406 - Orthorhombic - a 4.75800 - b 10.21200 - c 5.98800 - alpha 90.000 - beta 90.000 - gamma 90.000 - Primitive - Pbnm (62) - 4 - 290.949 - I/Ic PD
Operations: Import
d:\guests\StG-sand.RAW - File: StG-sand.RAW - Type: 2Th/Th locked - Start: 20.000 ° - End: 70.000 ° - Step: 0.020 ° - Step time: 1. s - Temp.: 25 °C (Room) - Time Started: 2 s - 2-Theta: 20.000 ° - Theta: 10.000 ° -
Lin
(C
ou
nts
)
0
100
200
300
400
500
2-Theta - Scale
20 30 40 50 60 70
42
Table 2.1 The ZB60 binder ingredients for the ZP131 material [49].
Ingredient Approximate amount (% mass)
Homectant 1 <10
Homectant 2 <8
Polymer <4
Water 85-95
Table 2.2 ZB58 binder composition for the ZCast501 material [50].
Ingredient Approximate amount (% mass)
Glycerol 1-10
Preservative (Sorbic acid salt) 0-2
Surfactant <1
Pigment <20
Water 85-95
2.2 Experimental Design for Mould Material Analysis
The mathematical model used to gauge responses to combinations of baking
time and temperatures is governed by statistical variance analysis. The
objective is to show the relationship of changes in the independent variables
of time and temperature to the outcomes (responses) MCS and MP.
Experiments proceeded by way of creating a matrix of varying time and
temperature combinations, followed by a regression analysis of these
outcomes. The results of the regression analysis for MCS and MP each
devolve to a second order polynomial expression with two independent
variables.
The experimental design used is a Central Composite Design (CCD) which
yields a classical quadratic equation. This design, shown below in Figure 2.3,
has sampling points at evenly spaced intervals from the central point. This is
undertaken on the basis that it is not known in advance how the response
surface will be orientated around the x-y axes and there is no presumption
that the standard error will be greater in some directions rather than others
43
[51]. This provides the maximum information on each response with fewer
trials than by using standard factorial experimental designs. This is important
as the mould materials are relatively expensive.
Figure 2.3 Data plot of the points showing the different sampling combinations of the
central composite experimental design.
Figure 2.3 shows temperatures (X1) and times (X2) at which the mould is
baked in a furnace. The CCD allows variation of baking time from 3.17-8.88
hours and temperature from 130-270°C, with the sampling points as shown, at
equal distances from the central point. Each variable is labelled with both
natural and coded values. For example 200°C is transformed into a coded
variable which in this case is 0 at X1.
The natural points of temperature and time, namely (130,6), (200,8.88),
(200,3.17) and (270,6) form the star points of the model. The star points are
those points at a distance α from the centre point and these points establish
new extremes for the factors. In a CCD design α is equal to n , where n is the
number of factors; in this case two. Thus α = 2 1.414 .
The CCD for outputs of MCS and MP proceeds in three separate parts. The
first part is a basic two level factorial segment of the design, which allows the
estimation of the linear terms [52].
44
For the factorial segment, the basic form is:
vn (2.1)
Where:
v = number of levels at which factors is varied
n = Number of factors; e.g. time and temperature
In this case, the linear portion is 22 as there are two factors, time and
temperature, and the factorial experiment only considers the maximum and
minimum of each factor, as shown in Table 2.3 below.
Table 2.3 The Factorial part of CCD
Levels
-1 +1
X1 (T,°C) 150 250
X2 (t, hrs) 4 8
The second segment of the CCD consists of the number of experimental
repetitions at the centre, which is at the zero level of each factor or in this
case (0,0) = (200°C,6 hrs). The third segment of the design consists of axial
or star points as shown in Table 2.4, which contribute to the estimation of the
quadratic portions of the model. These points are denoted distance α from the
centre. This distance is defined through the following equation.
1
4f(n ) (2.2)
Where: nf = number of experimental runs in the factorial portion.
This means that: 1
4f(n ) 2 1.414
(2.3)
45
Table 2.4 Experimental design table showing (a) the x matrix of the time and
temperatures as coded variables and (b) showing the design table with the natural and
coded variable combinations
2 20 1 2 1 2 1 2
1 1 1 1 1 1
1 1 1 1 1 1
1 1 1 1 1 1
1 1 1 1 1 1
1 2 0 2 0 0
1 2 0 2 0 0
1 0 2 0 2 0
1 0 2 0 2 0
1 0 0 0 0 0
1 0 0 0 0 0
1 0 0 0 0 0
1 0 0 0 0 0
1 0 0 0 0 0
x x x x x x x
(a) (b)
Table 2.5 The CCD experimental design with additional star points
Levels
Coded variable -1.414 -1 0 1 1.414
X1 (T°C) 130 150 200 250 270
X2 (t hrs) 3.17 4 6 8 8.88
This is translated into the X matrix and the design table, as shown in Table
2.4.
Actual experiments are conducted as per the design table, using the
combinations of different factors. Using the responses measured and the
matrix multiplication, polynomial expressions are developed for both MCS and
MP, in the following general form:
2 2
0 1 1 2 2 11 1 22 2 12 1 2y X X X X X X (2.4)
46
Where:
y = Is the response at X1 and X2 (for both MCS and MP with their respective
co-efficients).
2.2.1 ANOVA Calculations
The two equations for MCS and MP as per equation (2.4) must now be
analysed:
(A) To test the overall model significance in fitting the experimental data.
(B) To test the linear, quadratic and interaction contributions of the model to
results achieved.
First, the ANOVA examines the suitability of the regression analysis of the
model in terms of its accuracy and its confidence (A). Then, ANOVA tests the
significance of the linear and quadratic segments of the model and the overall
residual term (B). This is accomplished through the use of difference of sum
of squares.
Sum of Squares Regression Term (A)
n2
R i
i 1
ˆSS (y y ) (2.5)
Where:
ˆiy = The calculated result from the model at interval i
iy = Average of experimental results
Sum of Squares Residual Term
n2
RES i i
i 1
ˆSS (y y ) (2.6)
47
Where:
iy = The experimental result at interval i
ˆiy = The calculated result from the model at the interval i
Pure Error (PE) Term and Lack of Fit (LOF)
RES PE LOFSS SS SS (2.7)
n2
PE ij i
i 1
SS (y y ) (2.8)
Where:
ijy = Experimental result at centre at interval I of that group, j
iy = Average of calculated results from model of group j
Thus SSLOF = SSRes - SSPE
This describes the lack of fit of the model to the experimental data.
Sequential Sum of Squares (B)
This analysis establishes the effect of the linear, square and interaction terms
separately and quantifies their individual contribution to the overall residual
sum of squares. To establish these separate effects the model equations
need to re-fitted to the experimental values to create separate new linear,
quadratic and interactive equations. Thus, considering only the linear terms of
equation (2.9) the equation then becomes:
0 1 1 2 2y X X (2.9)
From here, the residual sum of squares is calculated using the following
equation.
48
n ^2
xx
1
(y y) (2.10)
Where:
yxx = Predicted value using equation (2.9)
^
y = Predicted value using entire model
So the difference between the case of the linear effect and quadratic effect is
shown below.
seq;linear 0 1 2 3 4SS ESS | (2.11)
Where: ESS = o 1 2
n ^2
, ,
1
(y y) -o 1 2 3 4
n ^2
, , , ,
1
(y y) (2.12)
The values in parentheses in equation (2.11) refers to the constants in the
separate linear and quadratic equations. Equation (2.12) shows that the
sequential sum of squares is equal to the difference of the error sum of
squares between linear terms and quadratic terms of the model. This process
is further repeated for the remaining part of the polynomial, i.e. the interaction
component2.
Degrees of freedom (DOF)
The DOF for the ANOVA is the number of independent parameters to
describe a particular component in a model, here the components being MCS
and MP. For the overall model there are six parameters, resulting in five DOF.
The linear, quadratic and interaction DOF are listed below, followed by the
other terms in the ANOVA table.
2 Appendix B details the relevant details.
49
DOF linear: Terms o 1 2, , form the linear part thus DOF = 3-1 = 2
DOF Quadratic: Terms o 11, 22, for the Quadratic part thus DOF =3-1 = 2
DOF Interaction: Terms 0, 12
for interaction part thus DOF = 2-1 =1
DOF Residual: This is the total DOF for both pure error and lack of fit and is in
the form of the equation:
RESDOF N P (2.13)
Where:
N = Total number of experiments
P = Number of parameters (constants) in equation
DOF Pure Error is:
PE replicateDOF (N 1) (2.14)
Where: Nreplicate = The number of experimental runs at the centre
DOF Lack of Fit: LOF RES PEDOF DOF DOF
(2.15)
F test
The F test or F statistic determines the significance of the amount of variance
explained by each of the linear, quadratic and interaction terms when
sequentially added to the model for determination of MCS and MP. The F test
is based on the following equation:
R Ro
RES RES
SS / DOFF
SS / DOF (2.16)
If this number, Fo is less than the number found from standard tables, the Null
Hypothesis is true and the significance of the model or lack of fit of the model
50
to the data is statistically due to chance. If Fo is higher than the standard
reference the null hypothesis is rejected and the model correlation is said to
be statistically significant at the chosen probability level. If a probability (P)
value is <0.1 (confidence level 1-P) then 0.1 (10%) is used and the null
hypothesis is rejected. This means that the there is a no greater chance of
10% of the result being due to chance.
2.3 Experimental Methodology for Testing Mould Materials
Experimental testing was conducted on 3D printed mould materials to
determine MSC and MP. As noted above, statistical design of experiments
was applied to ensure that the variability in the results was quantified in a
scientific manner. The other primary use of statistical experimental designs
was to reduce the number of experimental trials to decrease the consumption
of the 3D printer materials.
Mathematical models for critical mould properties were developed as above,
section 2.2. The ZCorp materials utilised were a plaster (ZP131) and plaster
sand mixture (ZCast501), with the latter supplied with guidelines for a post
print baking process. The general baking guidelines for ZCast501 is specified
as 4-8 hours and 200°C.
The critical characteristics of sand moulds are MCS and MP [53]. High MCS is
required to produce suitable moulds for metal casting. High MP allows relief
from gas pressure build up during casting. The baking times and
temperatures of the mould materials were varied to established their effects
on responses (MCS and MP) in a two factor central CCD.
After printing and baking, all printed mould specimens were kept in a
controlled low humid environment in a dessicator, shown below in Figure 2.4,
to ensure consistant low humidity of the specimens. This was done to remove
any complex influence that moisture may have on the MCS and MP and
ensure results are only due to the influences of varying time and temperature
of baking.
51
Figure 2.4 Photograph of the desiccator used to prevent moisture formation
All specimens were baked in electric furnaces at various temperatures and
times. Each furnace was tested with a CHY K-type thermocouple to ensure
that the temperature control was accurate to ±10°C. In all cases the curing
temperature was reached first before the samples were placed into the
furnace. Once the specimens had cooled down after baking, to room
temperature, they were placed in desiccators to minimise any moisture
effects. Testing was conducted on both mould materials used by the ZCorp
310 printer.
2.3.1 Mould Compressive Strength (MCS)
Compression tests on the mould samples were conducted using the
Hounsfield equipment at AUT University, shown below in Fig. 2.4, using a
length to diameter ratio of 2:1. The specimen shape was based on the ASTM
D7012 – 10 standard [54].
52
Figure 2.5 Photograph depicting the compressive test of the mould material.
The testing methods were those outlined in [54] which state that the loading
rate should be at such a rate that will cause the specimen to fail within 5-10
minutes. This corresponded to a loading rate of 0.1 mm/min. However it is
noted that compressive strength testing on rock like materials can compress
onto themselves after initial fracture, giving values higher than the true value.
To allow for this, each specimen was also observed visually for cracks at the
time of initial failure. The significant drop in strength and visual sign of a crack
were utilised as the true indicators of ultimate compressive failure.
2.3.2 Mould Permeability (MP)
Permeability measurements were made at the Geo-mechanics laboratory of
the Civil Engineering Department of the University of Auckland. Printed
cylindrical ZCast501 and ZP131 specimens, 78mm in length and 38mm in
diameter, were tested. Each specimen was applied with a silicone sealant
shown below in Figure 2.6a The sealant was applied around the outside
longitudinal surface of the cylinder and then fitted with an impermeable rubber
membrane. The specimen was then allowed to dry for 12 hours before being
placed in the load cell, with rubber rings placed on the top and bottom steel
seats. The role of the sealant was to prevent any air escaping outside the
cross section and around the specimen‟s outer surface. Once the specimens
were sealed, each specimen was contained in water to ensure that the rubber
53
membrane would not expand radially outwards. Any air passing over the outer
cylindrical surface of the specimen, rather than only through the cross section,
would result in inaccurate readings.
(a)
(b) (c)
Figure 2.6 The Permeability set up showing (a) the sealant used to adhere the
membrane to the specimens (b) rubber rings and specimen with membrane applied
and (c) load cell apparatus ready for testing
The pressure from the air tank was set at 16KPa. Once the pressure had
stabilised, an inline float type flow meter, located on the outlet side of the
specimen, was read every two minutes from zero to thirty minutes. The
permeability of materials changes over time but testings related to steady
state permeability values only. The flow rate of the passing fluid (air) was
subsequently plotted against time to establish when the permeability value
had stabilised. In most cases the flow rate stabilised after thirty minutes, but
54
some specimens stabilised after twenty minutes. Stable conditions were
assumed when six minutes had passed, with no further increase in the flow
rate. The flow readings were then coupled with atmospheric pressure data
sourced from the National Institute of Water and Atmospheric research
(NIWA) to calculate permeability values using Darcy‟s equation3. An Auckland
weather station was chosen as reference station, and instruments present in
the geo-mechanics lab were used to reference the ambient temperature and
relative humidity.
Figure 2.7 Photograph of the Tri-axial cell set up for the permeability testing
3 See Appendix A1
55
Figure 2.8 Close up of tri-axial cell test apparatus whilst testing
2.4 ZP131 Material Results and Discussion
2.4.1 MCS Model
Earlier casting trials by the author showed that ZP131 gives better results in
terms of cast surface roughness, dimensional stability and strength than
castings produced in ZCast501 moulds. As a result, it was decided to use
ZP131 as a moulding material, and a full analysis of the moulding
characteristics of this material was considered. The testing procedures
described above were used to analyse the variation of MCS and MP of this
material. The regression model developed for MCS are shown in equation
(2.17).
MCS
2(T,t) 1.64802 0.00025T 0.19267t 0.00001T
20.00151t 0.00091Tt (2.17)
The ANOVA below in Table 2.6 showed that the MCS regression analysis
was significant. The overall regression of the MCS was found to be
statistically significant at a 96.5% confidence level. The low P value for the
regression term meant that a significant portion of the total variance was
56
accounted for by the regression model and this was confirmed with a high R2
value, showing that 78% of the variation was explained by the regression
model. Thus an adequate representation of the overall relationship between
MCS and the temperature and time of baking was established. The fact that
no significant linear, square or interaction effects were found indicates that no
individual effect was significant, rather the combination of all of these. This
can also be observed in the response surface plot shown in Figure 2.11,
which shows the linear nature of the surface but also the curved square effect.
Also the surface plot revealed the optimum levels to be 150°C and 8 hours for
maximum MCS.
The pure error Sum of Squares (SS) term accounted for roughly two thirds of
the overall residual SS. This led to a low value for the lack of fit of SS,
showing that the model was fitting the data adequately. The large error term
reflects the deviations among the repeated centre points trials and shows that
there was a large standard deviation.
Table 2.5 ANOVA of compressive strength model of ZP131
Source DF Seq SS Adj SS Adj MS F P
Regression 5 1.71474 1.714745 0.342949 5.28 0.025
Linear 2 1.67845 0.019296 0.009648 0.15 0.865
Square 2 0.0033 0.003298 0.001649 0.03 0.975
Interaction 1 0.033 0.032995 0.032995 0.51 0.499
Residual Error 7 0.45438 0.454376 0.064911
Lack-of-Fit 3 0.11972 0.119724 0.039908 0.48 0.715
Pure Error 4 0.33465 0.334653 0.083663
Total 12 2.16912
R2 = 79.05%
57
Figure 2.9 Compressive strength Vs baking time and temperature for ZP131
Variation of MCS model in Figure 2.9 shows that the best results were
obtained at the lowest temperatures and highest time exposures. Increasing
the baking temperature resulted in an almost linear decrease in the MCS over
the entire time range, as shown in Figure 2.10a. By increasing the time
interval however, this linear relationship to strength remained, but the slope
gradually became more negative as time increased. This was most likely due
to the removal of moisture from the gypsum plaster particles. This increased
heat energy resulted in lower strength values, obviously affecting the bonding
strength of the gypsum material. Increased baking temperatures removed
water of crystallisation present in the gypsum microstructure [48], which
results in weakening the microstructural strength of the plaster.
With respect to varying time (Figure 2.10b), the effect of increasing the
temperature level was a decrease in the MCS response. However, there was
a small interaction effect, with increasing temperatures changing the slope of
the MCS expression from positive to negative over time. The ANOVA table
shows that this interaction effect was the most significant individual model
effect in the MCS analysis.
3
4
5
6
7
8 100
150
200
250
300
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Temp. (°C)
Ultimate Compressive Strenth (UCS)
Time (hours)
UC
S (
MP
a)
58
(a) MCS Vs Temperature
(b) MCS Vs Temperature
Figure 2.10 The Variation of Mould Compressive Strength of ZP131 with (a) baking
temperature and (b) baking time
0.25
0.75
1.25
1.75
2.25
120 150 180 210 240 270
UC
S (M
Pa)
Temperature (°C)
t = 3.3 hours t = 4 hours t = 5 hours
t = 6 hours t = 7 hours t = 8 hours
t = 8.8 hours
0
0.25
0.5
0.75
1
1.25
1.5
1.75
2
2.25
3 4 5 6 7 8 9
UC
S (M
Pa)
Time (hours)
T = 130 °C T = 150 °C T = 180 °C T = 200 °C
T = 230 °C T = 250 °C T = 270 °C
59
2.4.2 MP Model
The ANOVA of the MP model is given in Table 2.6 and revealed no
statistically significant model effects with respect to varying time and
temperature of baking. The coefficient of determination (R2) reveals that only
38% of the total variance was accounted for by the chosen factors. The R2
value was low, due to the large residual variance not explained by the model.
This variance was comprised of pure error and model lack of fit. The pure
error term was found as the majority of the total residual SS. This indicates
that there was considerable variation in the trials that were repeated under the
same conditions. The model lack of fit also accounted for just below half of
this overall residual, highlighting the inadequacy in modelling the data. The
model shows that the best MP values were achieved at low temperatures and
baking times, shown below in Figure 2.11. The response surface (Figure 2.11)
showed that increasing baking time and temperatures resulted in decreasing
MP of ZP131 samples, with an optimum point established at 150°C and 3.2
hours. The model however, was able to fit the data, with no statistical lack of
fit found. This all suggests that MP was not significantly affected by the time
and temperature baking process over the chosen intervals.
2 2MP T,t 6199.86 18.16T 635.65t 0.03T 41.78t 0.41Tt (2.18)
Table 2.6 ANOVA of permeability model of ZP131
Source DF Seq SS Adj SS Adj MS F P
Regression 5 414756 414756 82951 0.86 0.55
Linear 2 181393 189423 94712 0.98 0.421
Square 2 226773 226773 113387 1.18 0.363
Interaction 1 6590 6590 6590 0.07 0.801
Residual Error 7 674977 674977 96425
Lack-of-Fit 3 303285 303285 101095 1.09 0.45
Pure Error 4 371692 371692 92923
Total 12 1089734
R2 = 38.06%
60
Figure 2.11 Permeability Vs baking time and temperature in the case of ZP131
The permeability plots in Figure 2.12 below show that the material was found
to be more permeable at conditions closer to the green state, i.e. not baked.
The effect of increasing baking temperature, shown in Figure 2.12a, resulted
in a lower permeability curve. This trend continued at baking times from 3.3-6
hours but longer exposures from 7-8.8 hours resulted in a slight increase in
the permeability curve. Also little interaction effects were observed, with
changing factor levels resulting in a shift in the MP curve, rather than altering
the slope of the curve. This relationship was quantified in the above ANOVA
table with low P values for the interaction term.
With respect to time (Figure 2.12b), permeability decreased as time was
increased. However after around 7 hours, Figure 2.12b shows the MP began
to rise slightly. The lower permeability curves were unusual as it was
expected that the permeability would slowly increase as the moisture was
removed from the sample. It was possible that as the baking temperature was
increased bonding strength was lowered, which may have led to some
blocking of intergranular pathways. This would account for the initial
decreasing trend, with the prolonged heating effect removing some binder
constituents and slightly raising permeability.
3
4
5
6
7
8
100
150
200
250
300
1800
2000
2200
2400
2600
2800
3000
Time (hours)
Permeability
Temp. (°C)
Perm
eabili
ty (
mD
)
61
(a)
(b)
Figure 2.12 Permeability plots with respect to (a) varying temperature and (b) time
1800
2050
2300
2550
2800
100 150 200 250 300
Per
mea
bili
ty (
mill
iDar
cy's
)
Tempreature (°C)
t = 3.3 hours t = 4 hours t = 5 hours
t = 6 hours t = 7 hours t = 8 hours
t = 8.8 hours
1750
2000
2250
2500
2750
3000
2 3 4 5 6 7 8 9
Per
mea
bili
ty (
mill
iDar
cy's
)
Time (hours)
T = 130 °C T = 150 °C T = 180 °C T = 200 °C
T = 230 °C T = 250 °C T = 270 °C
62
2.5 ZCast501
2.5.1 ZCast MCS
The same analysis for ZCast501 was earlier undertaken by the author in the
final year of the undergraduate study. The response models, essential results
and discussion, are presented here for the sake of continuity. The regression
models for MP and MCS are given by equations 2.19 and 2.20 respectively.
2 2MP(T,t) 2962.12 10.94T 1231.25t 0.01T 57.6t 2.24Tt (2.19)
MCS
6 2(T,t) 3.535 0.03205T 0.694t 80 10 T
2 40.0514t 7.7 10 Tt (2.20)
The response surface for the MCS of the ZCast501 samples are shown in
Figure 2.13 and the ANOVA of the model is given in Table 2.7. The ANOVA
shown in Table 2.7 showed that all model parameters were significant with all
parameters above the 90% confidence level, with the exception of the
interaction term. The addition of the square term was the most significant with
a confidence of 94.8 percent. The linear SS term was also significant at 90.8
percent confidence.
The overall regression model had 96.8% confidence level and accounted for a
significant portion of the total variance in the experimental results. The model
also fitted the experimental data adequately. A small residual term meant that
most of the variance was accounted for by the varying baking time and
temperature factors. The repeated centre point trials in the experiment were
found to be consistent and close to the grand average of the experimental
results.
63
Table 2.7 ANOVA table of the MCS model
Source DF Seq SS Adj SS Adj MS Fratio P
Regression 5 1.00818 1.00818 0.20164 6.35 0.032
Linear 2 0.62379 0.25341 0.1267 3.99 0.092
Square 2 0.36092 0.36092 0.18046 5.68 0.052
Interaction 1 0.02348 0.02348 0.02348 0.74 0.429
Residual Error 5 0.15883 0.15883 0.03177
Lack-of-Fit 3 0.09757 0.09757 0.03252 1.06 0.519
Pure Error 2 0.06126 0.06126 0.03063
Total 10 1.16701
R2 = 86.39%
Figure 2.13 Mould Compressive Strength Vs time and temperature of baking for ZCast
501
At times ranging from about 3-6 hours, the MCS was seen to increase with
temperature, reach a maximum and then sharply fall with further time and
temperature exposure. The initial rise in the compressive strength is
substantial and in most cases about 0.2 MPa. This behaviour was attributed
to the temperature acting as a catalyst on the liquid binder.
3
4
5
6
7
8 120140
160180
200220
240260
280
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Temp. (°C)
Ultimate Compressive Strength (UCS)
Time (hours)
UC
S (
MP
a)
64
From Scanning Electron Microscopy (SEM) work, conducted at the
Department of Environmental Science of University of Auckland, high
definition and high magnification pictures of the fractured specimens, from the
MCS, were able to be collated. The liquid binder forms a bond between the
two different materials, i.e. the gypsum and the olivine sand. This curing of the
binder subsequently strengthens intermolecular bonds, leading to the initial
rise in compressive strength.
Figure 2.14 ZCast501 mould sample baking temperature was 150°C and time equal to 4
hours
The SEM photograph shown above in Figure 2.14 shows the gypsum
particles adhering well to the larger sand grains, acting as a solid binding
material in the overall structure. This better adhesive quality of the gypsum
was seen at the lower temperatures and thus provided a strengthening
structure, producing more resistance to the dislocation movement of the larger
sand particles.
With further heating, however, the MCS was seen to drop rapidly after
variable changes to time temperature above 6 hours and 200°C respectively.
This was possibly caused by the removal of the water from the gypsum
material, as water is vital for the gypsum to retain its crystal structure and
strength. High temperatures and baking times produced severe dehydration of
65
the gypsum particles, as shown in the SEM photographs (Figure 2.15). This
resulted in highly disorganised and weakened gypsum. The liquid (mainly
water) binder present in the gypsum was effectively removed and as expected
the MCS reduced dramatically.
Figure 2.15 SEM photograph showing the embrittled gypsum at the grain boundary
As a result, the gypsum became embrittled and upon application of load the
brittle gypsum at the grain boundaries was easily broken and classic
intergranular failure resulted. The force of the dislocating sand grains
fractured the relatively weak brittle gypsum.
This sudden fall of strength could explain some of the variability in the results.
Also, the mechanics of aggregates is not an exact science and their behaviour
under load is much more difficult to quantify than for metals. Furthermore,
there are two distinctly different materials present in one bulk structure. This
means that the overall material is complex. The results of MCS testing show
large variations for experiments conducted at the same temperature and
baking time, indicating a high sensitivity to the consistency of samples. It
Embrittled
gypsum
Large sand
grain
66
appears there may be a dramatic impact on MCS results from small changes
to the composition and proportions of the sample materials. This conclusion
cross-references with the large material grain size distribution found by the
author during the undergraduate work.
2.5.2 ZCast MP
The MP response surface model shown bellow in Figure 2.16, allowed the
estimation of the optimum temperature and time of baking as 209.3°C and
7.70 hours respectively for the maximum MP. Table 2.8 below shows the
ANOVA conducted for the permeability model, which showed significant
reductions of the residual SS in the linear, square and interaction portions of
the model. These reductions were all significant at >90% confidence,
suggesting that each term reduced the residual SS. The interaction effect
showed that there was a positive impact on permeability when temperature
was held constant and time was varied at different levels (and vice-versa).
Table 2.8 The ANOVA table of the permeability model
Source DF Seq SS Adj SS Adj MS Fratio P
Regression 5 856932 856932 171386 3.37 0.072
Linear 2 259770 685311 342656 6.73 0.023
Square 2 396040 396040 198020 3.89 0.073
Interaction 1 201121 201121 201121 3.95 0.087
Residual Error 7 356384 356384 50912
Lack-of-Fit 3 7067 7067 2356 0.03 0.993
Pure Error 4 349317 349317 87329
Total 12 1213316
R2 = 70.63
With reference to the surface plot, the MP of the ZCast material increased
with time from 3.3-6 hours. It is conjectured that the initial heat energy
removed some of the binder material in the mould by evaporation. This initial
drying out or possible burn off of some of the minor constituents would result
67
in an increase in permeability, as the dynamic viscosity of the passing air
increased with moisture reduction, allowing easier access of air through the
material. Further, this initial heat energy was acting as a catalyst on the binder
present in the mould material. The smaller Gypsum material that bonds to the
larger sand grains acts as the glue in the bulk of the material. If the binder is
subjected to heat energy the catalyst effect of the temperature will cause a
stronger structure to be developed as the inter material bonds are
strengthened. This would mean increasing more intergranular pathways,
increasing MP.
Figure 2.16 Permeability Vs baking time and temperature in the case of ZCast501
SEM photography in Figure 2.17 below shows specimens which were baked
at lower times and temperatures. It is noted that the fine needle like gypsum
particles adhere well to the sand particles at these lower heating intensities.
Upon further heating, however, from 6-8 hours, the permeability reached a
plateau and then started to decrease as shown in Figure 2.17b. This sharp
reversal in trend was thought to be due to some of the material melting, fusing
or changing phase. This would mean that intergranular pathways would
become blocked or partially blocked, thus decreasing permeability.
3
4
5
6
7
8
100
150
200
250
300
800
1000
1200
1400
1600
1800
2000
2200
2400
2600
Time (hours)
Permeability (mD)
Temp. (°C)
Perm
eabili
ty (
mill
iDarc
ys)
68
(a) (b)
Figure 2.17 SEM photographs showing the gypsum structure of a mould baked at (a)
150°C and 4 hours and (b) modified structure of a mould baked at 270°C for 6 hours
Figure 2.17 confirms the significant change in the gypsum structure on
fractured surfaces of ZCast specimens. At high baking times and
temperatures (Figure 2.17b) the gypsum structure was loose and its bonding
strength reduced, resulting in a collapse of the intergranular pathways, thus
blocking the passageways to flowing air.
2.5.3 A Comparison Between the Two Materials
The regression models constructed for both the ZCast501 and ZP131
materials differed in that the ZCast501 regression models had more
significant coefficients that were able to explain a significant portion of the
total variance from the experimental results.
In both the ZCast models the majority of model effects were significant, with
considerable reductions in the SS of the residuals when each term was added
to the model. The MCS and MP responses were able to be adequately
modelled with the quadratic polynomial expression created by the CCD. The
effect of the time and temperature of baking had significant linear and square
components for both MCS and MP, with a significant interaction contribution
effect to MP.
69
The MCS ZP131 regression model was able to explain a significant portion of
the total experimental variance. However, in the MP ZP131 regression model
there was no statistically significant relationships found. There were no
significant reductions in the linear, square or interaction residual SS when
added to the model. Nevertheless, the overall fit of the models was found to
be adequate, with no significant lack of fit. The inability of the model to
account for MP variance indicates that the ZP131 material was not affected by
the temperature and time of baking unlike the ZCast501 material.
With respect to the actual MCS and MP values obtained, the two material
systems were similar in magnitude. The repeated centre points in the
experimental design (T=200°C and t=6hours) were the basis of comparisons
of the MCS and MP and the results are presented below. The MCS for ZP131
was higher than for ZCast501 material. This result was predicted as the
material has a finer and more uniform particle distribution, providing a more
homogenous structure. The difference in the average MCS was confirmed
with a T-Test between the ZCast501 and ZP131 material. The P value from
the T-test was 0.2678. This value was not significant at generally accepted
confidence intervals (P<0.1).
ZCast501 material was found to produce higher MP values. The mean
permeability of the ZCast501 and ZP131 materials was 2116.77mD and
2099.1mD respectively, (based on centre point trials). The magnitude of the
average permeability and the standard deviation were unexpectedly similar. It
was expected that the much more closely packed gypsum material in ZP131
would produce lower MP values. Similarly, it was thought that the ZCast
material, being both more porous and of larger average grain size, would
produce larger MP values. However the T-Test on permeability showed a high
P value of 0.9914, which meant that the two population variances were not
statistically different.
From the statistical testing for MCS and MP, using the two materials, it is clear
that no significant differences were found. However, on a practical basis
ZP131 seemed stronger and showed superior definition regarding the mould
70
faces and edges. Also the ZP131 material had smoother surface finish, which
was reflected in the initial castings involving the plaster material.
71
CHAPTER 3 CHARACTERISTICS OF STATIC CASTINGS
PRODUCED IN 3D PRINTED MOULDS
3.1 Casting in Printed Moulds
Following evaluation of 3D printed materials in Chapter 2, the next procedure
was to conduct casting trials to establish the suitability of these moulds for
casting light metals such as Aluminium (Al) and Magnesium (Mg). As little
work has been undertaken in casting light metals into these moulds, scant
information is available on the effects of varying the constituent parameters.
Typical experimental parameters are mould materials, mould coatings, alloys
cast and process parameters such as pouring temperature. However, with the
simultaneous variation of all these factors, an experimental design was
required to establish in a statistical manner, the effects of these parameters,
both individually, and in combinations, on the mechanical characteristics of
the castings. These trials were also aimed at attaining practical experience
with rapid casting of Al and Mg alloys. The following sections present the
experimental design used, together with the relevant mould design used in the
static casting trials.
Using the optimum mould material conditions developed in Chapter 2, the
main aim of these static casting trials was to establish the combination of
process variables which would yield the optimum mechanical responses for
light metal casting in 3D printed moulds. The properties of the castings
produced from these moulds would then be studied and compared to each
other and to traditional sand casting values. Many other mechanical
properties could have been chosen but given time and other constraints, the
following mechanical variables were evaluated and studied by analysing the
cast specimens:
72
Properties to be measured
Surface roughness
Ultimate Tensile strength (UTS)
Percent elongation
Brinell Hardness (HB)
Microstructural examination
Surface Roughness
The surface roughness of a casting is an indicator of the quality of the cast
surface produced from the 3D printed moulds. Surface quality needs to be
high as this reduces post production machining. Cast specimens also need to
have low surface roughness to ensure that good surface contact is able to be
made in joining applications.
Ultimate Tensile Strength
Knowledge of the strength is one of the most important properties of a
material, which has to function inside this limit to avoid failure. Also, the
tensile strength gives a good indication of the overall quality of the material
and its subsequent properties such as elastic limit or yield strength, and
modulus of elasticity.
Hardness
The hardness gives the measurement of how resistant a material is to
permanent deformation. This allows comparison of material hardness to
established values found in the literature.
73
Microstructural Examination
Photomicrographic analysis allows the establishment of any similarities or
differences between cast microstructures. Also the actual results of the tensile
strength and microstructure can then be connected by carefully studying the
microstructure of each specimen. These effects can also be quantified by
visual examination of the grains and grain counting.
3.2 Experimental Plan
Factorial or fractional factorial design of experiments, as used in Chapter 2,
could not be used here. There were two reasons for this: firstly, it is difficult to
model qualitative factors, e.g. mould coatings, in traditional design of
experiments. Secondly, the static casting trials required the variation of four
factors each at three separate levels. This number of factors and levels would
require too many trials if a traditional experimental design was used.
To achieve an efficient experimental design a Taguchi L9 Orthogonal Array
(OA) was chosen. The Taguchi L9 technique allows up to four factors to be
varied at three levels of each factor. The result is only nine experimental trials
compared to eighty-one if a traditional factorial design was to be utilised. The
four factors chosen were:
Mould material
Mould coating
Alloy type
Pouring temperature
Mould Materials
The effect of the mould type and coating used in metal casting is a primary
factor affecting the quality of the cast parts. The mould itself needs to be
capable of processing reactive hot metals and should possess suitable
strength and permeability. The mould materials used were the two 3D printed
74
mould materials (ZP131 and ZCast501) and a resin bonded Silica foundry
sand.
Mould Coating
The application of a mould coating to the mould surface will also have an
effect on the cast surface roughness and the nature of the mould metal
reaction. Three mould coatings with different base ingredients were utilised:
Graphite (Isomol 200), Zirconium (Zircoat) and Magnesium Oxide (Magcoat).
Alloy Type
The alloys used in these trials were varied to establish the suitability of these
moulds to process different light metals, namely Al and Mg alloys. The Mg
alloys4 chosen were AZ91HP (Al9%Zn1%, Mg base) and AM-SC1 (SC1)
(Nd1.9%Ce0.7%Zr0.5%, Mg base). Firstly, the AZ91HP (High Purity) alloy
was chosen primarily as it is a widely used alloy grade and is suitable for sand
casting applications [55]. The SC1 alloy is a newly developed sand cast alloy
aimed at being used in the production of automotive components, such as
engine blocks [56]. The Al grade used was A356 (Si7%Mg0.4%, Al base),
which is a common sand casting alloy used in a variety of applications [56].
Pouring Temperature
The temperature at which these metals are poured was tested over the range
of 690-770°C, as specified by relevant literature [53, 55, 57]. The pouring
temperature used provides data on the metal fluidity and mould metal
reaction, especially when using reactive alloy such as Mg.
Table 3.1 below shows the completed L9 OA for the static casting trials,
followed by the relevant levels of each factor, shown in Table 3.2. In Table
3.1, the four factors used in the static casting trials are displayed in each
4 Refer to Appendix B for alloy composition
75
column, with the actual level of each factor shown in the row below. The
actual experimental work commences with the chosen combinations of the
levels at each experimental trial.
Table 3.1 Taguchi L9 experimental design table with natural variables
Trail Mould
Coating Alloy
type
Temp
(°C)
1 ZP131 ISOMOL AZ91HP 690
2 ZP131 ZIRCOAT SC1 730
3 ZP131 MAGCOAT A356 770
4 ZCAST ISOMOL SC1 770
5 ZCAST ZIRCOAT A356 690
6 ZCAST MAGCOAT AZ91HP 730
7 SILICA ISOMOL A356 730
8 SILICA ZIRCOAT AZ91HP 770
9 SILICA MAGCOAT SC1 690
Table 3.2 Factor and level combinations for static casting trials
Factors
Mould
material
Mould
coating
Pouring
temperature
Alloy
Level 1 ZP131 Isomol 200
(graphite
based)
690°C Mg : AZ91HP
Level 2 ZCast501 Zircoat W
(Zirconium
Silicate based)
730°C Mg: AM-SC1
Level 3 Resin
bonded
Silica sand
Magcoat S
(Magnesium
Oxide based)
770°C Al: A356
76
3.2.1 Taguchi Response and ANOVA5
The Taguchi methodology proceeds mainly by analysis of the Signal to Noise
(S/N) ratio. This measures the robustness of a specific process, i.e. the
sensitivity of a process to changes in the chosen factor levels. The S/N ratio is
able to measure the amount of variation due to uncontrolled (noise) factors.
Therefore the S/N ratio is a variance indicator of the power of the signal
(strength of response) to the noise (changes in process factors). This
technique ensures both signal optimisation and minimum variation around the
chosen signal level [58]. The S/N ratio is the transformation of the Mean
Squared Deviation (MSD) which measures both the average and standard
deviation. Depending on the application (desired signal level), the following
formulas may be used when processing the S/N ratio:
‘Higher the better’
The higher the better equation shown below in equation (3.1), is used when
the response is desired to be as high as possible.
r
HB 10 2i 1 i
1 1S / N 10log ( )
r y
(3.1)
Where:
S/NHB = Higher is Better (HB) Signal to noise ratio
yi = Experimental response value
r = Response repetition
‘Lower the better’
If a process response is desired to be reduced to a minimum (e.g. surface
roughness) the following equation is used:
5 For specific calculations of terms in the ANOVA table refer to Appendix A
77
r2
LB 10 i
i 1
S / N 10log ( y ) (3.2)
Where:
S/NLB = Lower is Better (LB) Signal to noise ratio
yi = Experimental response
r = Repeated response
In all cases the greater the S/N ratio, the lower the variation of the measured
variable is around the target value [58].
3.2.2 Pooling of Factors
Pooling of experimental factors means assigning insignificant factors utilised
as the experimental error in the experiment. This can be accomplished by
different methods and specific criteria which avoid arbitrarily pooling factors.
Pooling is generally done to increase the error sum of squares by pooling an
insignificant factor. Using Taguchi OA, such as in L9 static casting trials, all
the column (factor effects) are occupied, meaning that the error DOF is zero,
which then disallows the calculation of F ratios, variance and confidence
levels. To create an error term, small column effects are pooled together to
create a larger error term (known as pooling up strategy) [58]. This technique
of pooling was used in the current experiments, with factors pooled if the
contribution of an effect was around 10% or less of the total variance present
in the experiment.
3.3 Mould Design
A mould to contain the molten metal was created, so that cast specimens
could be subsequently machined and tested. Mould design is a highly
differentiated area of expertise and sound mould design often requires
significant practical experience of what works and what does not. Much of the
current mould design knowledge utilised some empirical geometric
relationships and experiences from prior testing. A simple mould design was
78
considered, comprising of a top gated inlet, cylindrical specimens (the final
casting) and a feeder situated at the end of the cylindrical cavities to allow for
metal shrinkage upon solidification. The mould itself was horizontally parted to
form two halves, the upper half (cope) and lower half (drag) The mould halves
were located together by aligning studs located at each corner of the mould
and when casting a central clamp was applied to prevent any movement due
to the buoyant forces of the molten metal. Along the length of the cylindrical
specimens some vent holes were also added to improve permeability and
help release any pressure built-up inside the cavity.
The actual casting samples produced in these moulds consisted of four 15mm
diameter cylindrical specimens, to be machined down to produce tensile test
specimens6 for testing. The mould also consists of a ceramic pouring cup
which was placed at the inlet to provide better metallostatic pressure. At the
bottom of the pouring cup was also a ceramic foam filter with ten Pores Per
linear Inch (PPI), to help remove any impurities and oxides in the metal, and
also to laminate the metal flow.
(a) (b)
Figure 3.1 CAD drawings of the initial static mould design, showing (a) Bottom half of
the mould and (b) 3D view of completed mould design.
6 Refer to Appendix B for tensile test part dimensions.
79
Figure 3.2 Sectional CAD drawing initial mould design, showing pouring entrance cup,
cylindrical cavity and riser
3.4 Static Casting Experiments
The casting trials were conducted using AUT‟s induction furnace. A custom
made 1040 carbon steel crucible was constructed to fit the furnace
dimensions to contain the molten Mg alloys (AZ91HP and AM-SC1) due to the
severe reactions when in contact with refractory crucibles. The same crucible
was also used to heat the A356 Al. For each trial the melt temperature was
read off a CHK500 K-type thermocouple just prior to pouring. Each 3D printed
mould was baked at the optimum time and temperature. Prior to baking a
mould dressing was applied with a brush over the entire cavity. For a general
comparison castings were produced in silica sand bonded by an ester
hardened alkaline phenol-formaldehyde polymer resin.
The casting of the metals involved transporting the steel crucible and carefully
poring directly into the printed moulds. In the case of Mg castings, the molten
Mg was covered by gas mixture to purge out any potentially reactive oxygen.
A mixture of CO2 and refrigerant r134a was used in conjunction with Foseco
MAGREX 36 flux to prevent any violent oxidation once the metal was heated
to its melting point.
80
Figure 3.3 Photographs showing AUT University induction furnace setup (left) during
casting of Mg alloy (right)
A small quantity of flux was applied to the crucible prior to charging the
crucible with the alloy. In all melting trials, alloy in the form of commercially
pure ingots was melted to ensure the melt was of a high quality. The molten
metal was filtrated with the use of a 10PPi ceramic foam filter placed near the
bottom of the ceramic pouring cup. When casting Al, the melt was degassed
with Nitrogen with a degassing lance shown in Figure 3.4. Degassing was
performed for about 3-5 minutes for each melt. The pouring cup was also
coated with the mould dressing to avoid any reactions with the molten Mg.
Figure 3.4 Degassing lance used in casting trials
3.4.1 Mechanical Testing of Castings
The cylindrical cast specimen surface was tested for surface roughness. This
was conducted at AUT University with a Taylor Hobson Form Talysurf50
surface testing machine, shown below in Figure 3.5.
81
Figure 3.5 Taylor Hobson Talysur50 surface testing machine
Each tensile test specimen was tested on an arbitrary side of the cylindrical
specimen in the longitudinal direction and rested on finely ground, level V
blocks. A Gaussian filter was used to filter out any waviness effects of the
surface profile and the average surface roughness, Ra was found. On the
chosen testing side, three readings were taken to obtain an overall average
surface roughness. Suitable cut offs of 8mm and 0.8mm were used in
conjunction with the surface testing standard ISO 4288-1996.
Once each casting was completed, the riser and sprue sections were cut off
with a hacksaw and each cylindrical specimen was turned down to form the
tensile test specimen7. This was conducted at low speeds and feed rates to
avoid excessive heating of the alloy. Whilst turning the specimens, coolant
was applied to prevent excessive heating and local changes in the
microstructure. The tensile testing was carried out on the Hounsfield Tensile
testing equipment at AUT University shown below in Figure 3.6. The loading
rate was 1 mm/min in all cases. The diameter of the critical cross-section was
measured with a micrometer to obtain an accurate cross–sectional area.
7 CAD drawing of specimens is detailed in Figure B.10, Appendix B.
82
Figure 3.6 Photographs of the tensile testing performed at AUT with (a) showing the
testing machine set up and (b) showing close up the stressing of a tensile specimen
Sprue and riser sections were then machined flat on a milling machine to
prepare specimens for hardness testing. Samples were tested in accordance
to AS 1816.1-2007 on an Avery Brinell hardness tester with 10mm indenter at
AUT University (Figure 3.7). The indentation diameter resulting from the
hardness tests was measured with a hand held optical measurement device.
Figure 3.7 Brinell Hardness testing apparatus
83
After the mechanical testing, the transverse sections of the fractured tensile
specimens were examined with SEM at Auckland University to determine the
cause and method of fracture. Further light microscopic analysis was
conducted at AUT University metal research laboratory. A transverse section
was cut with a hacksaw at the midpoint of the tensile test specimen. After
subsequent filing to bring the surface smooth and level, each specimen was
mounted in a Struers LaboPress mounting machine at a temperature of 150°C
and with a force of 30 KN. Once this was competed each specimen was
sanded on a Metaserv rotary grinder on the following sand grades; 180, 320,
500, 800, 1200, 2500 grit. Water was used to lubricate and prevent heating of
the specimen surface and after each specimen was completed it was finely
polished on a Metaserv universal polisher with the use of a 0.2µm cloth and
1µm diamond paste. To provide lubrication a Fumed Silica lubricant was
used. After polishing the specimens were then etched with Wrecks reagent8
[59] in the case of the Al castings for 15 seconds and then dried. The Mg
castings were etched in a Glycol based etchant9 [60]. The mounted
specimens were submerged for around 5 seconds, then washed in water and
alcohol and then dried. The photomicrography was done on optical
microscope at magnifications of X50, X100 and X400. The microstructure was
characterised by grain counting to determine the average number in
accordance with ASTM E 112 – 96 [61]. The Planimetric method was used to
determine the number of grains within a given area. In this case the area was
a rectangular photomicrograph, which was measured with a calibration
measurement indicator.
3.5 Static Casting Results and Discussions
3.5.1 Surface Roughness
With the exception of the mould material factor, the ANOVA table (Table 3.3)
showed that the variance in the other factors accounted for only a small
8 See Appendix B for composition.
9 Refer to Appendix B for composition.
84
proportion of the total variation in surface roughness. The mould material was
found to account for over two thirds of the total variance in the experiment and
at a 95% confidence level.
Table 3.3 ANOVA table of the surface roughness response
Rank Source DOF SS Variance FRatio P value Percent contribution
1 Mould 2 68.43 34.21 26.33 0.04 68.97
2 Coating 2 16.12 8.06 6.20 0.14 16.25
3 Alloy 2 12.07 6.03 4.64 0.18 12.16
4 Pool Temp 2 2.60 1.30
2.62
Total 8 99.22
100.00
Upon examining the S/N ratios in Figure 3.8 below, the ZP131 mould material
was found to give the highest ranking level of the mould material factor. This
was followed by the Silica foundry mould and the ZCast mould respectively.
Figure 3.8 S/N ratio of mould material in the surface roughness results
The low surface roughness values achieved by the ZP131 moulds can be
explained by considering the small size of the gypsum particles and their
uniform grain size distribution. This meant that the mould surface had a lower
roughness than the other trialled moulds, which was reflected in the castings.
The presence of smaller gypsum grains decreases the mould roughness
-17.39
-23.93
-22.12
-25.00
-23.00
-21.00
-19.00
-17.00
-15.00
ZP131 ZCAST501 Silica Sand
S/N
rat
io
Mould material
85
profile, resulting in lower peaks and valleys on the mould surface. Further, the
dimensional form was enhanced, providing a sharp reproduction of the overall
shape and dimensions.
The highest values of surface roughness were observed in the ZCast501
moulds. The poor surface finish of the castings produced in the ZCast moulds
was a combination of the inherent mould properties. Primarily this was due to
the poor grain size and the shape and distribution of the ZCast mould
material. Previous work by the author, noted in Chapter 2, showed that there
was a large grain size distribution in the ZCast material, and disproportionate
larger sand grains in relation to the gypsum particles. To compound matters,
the mould surface was easily eroded due to the low bonding strength present
in the ZCast moulds. Also, the degree of spherical curvature of the added
Olivine sand, which is basically a crushed rock, [62] was detrimental to the as-
cast surface roughness. As a result, the grains were highly angular, which
created greater intergranular cavities where the metal could penetrate [55].
Figure 3.9 The SEM close up photograph of an angular Olivine sand grain present in
the material
SEM work showed highly angular grains, as indicated in Figure 3.9. Also from
this photograph, the difference in size of the Olivine sand grains relative to the
surrounding gypsum was evident. This size difference leads to a decrease in
86
surface quality, even with the application of the mould coating. Further, the
mould surface was quite fragile with surface grains easily dislodged. When
applying the mould coating this led to relatively poor surface quality before
metal was cast, as the act of brushing removed the surface particles.
The confidence levels of the mould coating parameter in Table 3.3 was almost
significant with a P value of 0.14, which was slightly higher than commonly
accepted confidence levels (P value <0.1). When considering the S/N ratios of
the individual mould coating levels, shown below in Figure 3.10a, it was found
that Magnesium Oxide (MgO) based Magcoat product was the best suited
coating in achieving the minimum surface roughness. It was noted that Mg
castings produced rougher surfaces, due to mould metal reaction, with Figure
3.10b showing that the two Mg alloy grades resulted in the lowest S/N ratios,
with the less reactive Al alloy resulting in the highest S/N ratio (best surface
finish).
(a) Mould coating (b) Alloy type
Figure 3.10 S/N ratios of (a) Mould coating and (b) Alloy type levels in the surface
roughness response.
The use of the different alloys in these experiments resulted in a P value of
0.18 slightly higher than the mould coating (0.14) and also outside the range
of accepted P values. The reactive nature of the Mg alloys on surface
roughness was significant. The propensity of Mg alloys to oxidise is governed
by the Gibbs free energy change of [63].When the alloy changes phase into a
liquid, the alloy begins to oxidise and revert to its stable oxide form. The
application of a mould coating can reduce this reaction by providing a stable
-23.02
-20.47
-19.96
-25.00
-23.00
-21.00
-19.00
ISOMOL ZIRCOAT MAGCOAT
S/N
rat
io
Mould coating
-21.86
-22.07
-19.52
-23.00
-22.00
-21.00
-20.00
-19.00
AZ91HP SC1 A356
S/N
rat
io
Alloy type
87
exterior material, which facilitates a decreased ability for the alloy to oxidise.
The MgO coating proved to be a very stable coating material, relatively
unaffected by the molten Mg due to its large Gibbs free energy change.
The pouring temperature was shown to have a very insignificant effect,
leading it to be pooled as the error term. This result indicated that the effect of
pouring temperature did not affect surface roughness. Increased
temperatures over this range might have been expected to produce higher
surface roughness due to increased reaction but this was not observed. It is
noted, however, that this outcome was only over a total temperature variation
of 80°C. Moreover, temperature measurement was restricted to a
thermocouple probe in an open crucible. These qualifications could produce
measurement error of perhaps 10-20%, but this is not considered significant
in assigning pouring temperature as the experimental error.
Figure 3.11 S/N ratio of the pouring temperature factor levels on the surface roughness
values
All experimental results are presented in Table 3.7, while the overall rankings
of the factors are listed in Table 3.8. Cast specimens surface roughness were
23.6μm for AZ91, 11.4μm for A356 and 14.33μm for SC1 for castings
produced in the ZCast moulds. Surface roughness testing conducted on the
castings produced by ZP131 moulds showed better surface finishes, due to
the smooth surfaces of the ZP131 moulds. This was because ZP131 was not
-20.72-20.82
-21.91
-23.00
-22.00
-21.00
-20.00
690°C 730°C 770°C
S/N
rat
io
Pour Temperature
88
mixed with any foreign material. The surface roughness values were seen in
the ZP131, with 9.5μm for AZ91, 7.3μm for A356 and 5.8μm for SC1. These
values were acceptable with typical sand cast surface roughness values
ranging from 6-13μm [64].
Table 3.4 Static L9 experimental design table
Factors Responses
Mould
Coating Alloy
type
Temp
(°C)
UTS
(MPa)
Strain
%
Ra
(μm)
HB
ZP131 ISOMOL AZ91 690 156.42 2.17 9.49 56.82
ZP131 ZIRCOAT SC1 730 130.02 2.56 7.33 48.17
ZP131 MAGCOAT A356 770 127.13 0.75 5.84 60.54
ZCAST ISOMOL SC1 770 127.99 2.28 23.66 48.87
ZCAST ZIRCOAT A356 690 132.05 0.85 11.47 65.70
ZCAST MAGCOAT AZ91 730 150.87 1.52 14.33 60.54
SILICA ISOMOL A356 730 93.27 0.73 12.63 52.63
SILICA ZIRCOAT AZ91 770 127.11 1.54 13.98 56.82
SILICA MAGCOAT SC1 690 157.39 3.36 11.79 47.48
Table 3.5 Overall ranking of factors and their level from static casting trials
Factors
Response Mould
material
Mould
coating
Alloy type Pouring
temp.
UTS (ZP131) :4th (Magcoat):3rd (AZ91HP):1st (690°):2nd
Strain (ZP131):3rd (Magcoat) :4th SC1:
99.99%:1st
690°C:90%:
2nd
Ra ZP131:95%:
1st
(Magcoat) :2nd (A356):3rd (690°C):4th
HB (ZCast501):2
nd
(Zircoat):3rd A356: 90%:1st (690°):4th
* Brackets denote highest ranked level of non-significant factors
89
3.5.2 Ultimate Tensile Strength (UTS)
The ANOVA in Table 3.6 showed no significant factors at any of the tested
confidence levels. This result indicates that no single factor from the chosen
variables was influencing the UTS and the variation observed is due to a
combination of effects of all constituent factors. The ANOVA shows that the
total variance was comprised of similar magnitudes, with the mould coating
contributing 19.8%, the alloy 38.77% and the pouring temperature 30.21%.
The mould material was seen to have a very small effect (11.21%) and was
pooled as the error term. The negligible effect of the mould material perhaps
is due to a minimal variation in the cooling rates, as both moulds are coated
with the same materials.
Table 3.6 ANOVA of UTS response
Rank Source DOF SS Variance FRatio P value Percent contribution
1 Alloy 2.00 5.92 2.96 3.46 0.22 38.77
2 Temp 2.00 4.61 2.30 2.69 0.27 30.21
3 Coating 2.00 3.02 1.51 1.77 0.36 19.80
4 Pooled
Mould 2.00 1.71 0.86
11.21
Total 8.00 15.26 1.91
100.00
The alloy type accounted for the largest variations in the experimental results.
However, the end results were somewhat skewed because the UTS is directly
related to the alloy system itself. In the present case, the Mg alloy AZ91HP
scored better than the others as seen from the S/N ratios presented in Figure
3.12. The Al alloy was found to have the lowest strength values. The probable
reason for this was the presence of hydrogen porosity, which weakened the
as-cast structure. These aspects are covered more closely in section 3.5.2.
90
Figure 3.12 S/N ratio of the Alloy type factor levels in the UTS response
The second ranking factor in the ANOVA table was the pouring temperature,
which returned a P value of 0.27. This result was perhaps linked to the
reduction in the solidification time when lower casting temperatures were
utilised. An examination of the S/N ratio of this factor, in Figure 3.13, shows
that the best UTS can be obtained at the lowest temperature, 690°C. The
reduction in the solidification time in theory should result in smaller dendrite
grain sizes and a more refined grain structure. Coarser grain structures are
particular important in Mg and Al alloys for the UTS of the cast material [58,
65-67].
Figure 3.13 S/N ratio of the Pouring temperature Factor levels in relation to the UTS
response.
43.18
42.79
41.3041.00
41.40
41.80
42.20
42.60
43.00
43.40
AZ91HP SC1 A356S/
N r
atio
Alloy type
43.41
41.75
42.10
41.6041.8042.0042.2042.4042.6042.8043.0043.2043.4043.60
690°C 730°C 770°C
S/N
rat
io
Pour temperature
91
The significance of the mould coating was not expected considering the small
layer thickness. When considering the S/N ratios of mould coatings, the
Magcoat product performed much better than the other two coatings trialled.
The Magcoat product also gave the optimum surface roughness values, as
noted above, indicating that lower surface roughness values may be linked to
the better than expected UTS response of Magcoat.
Figure 3.14 S/N ratio of the mould coating factor levels in relation to the UTS response.
The mould material was found to be the least significant factor in achieving
the maximum UTS values. The thermal conductivity of plaster material was
much lower than that of the Olivine sand [68]. It was thought that a casting
produced in a ZCast501 mould may have slightly improved mechanical
properties compared to that from a mould made of plaster (ZP131) only. This
reflects the ability of ZCAST material to transfer heat at a higher rate, allowing
the casting to solidify faster and permit a finer grain structure. However, S/N
ratio analysis in Figure 3.15 below, showed that the ZCast mould material had
a negligible effect and actually the best mould in terms of strength was the
ZP131 plaster mould. It was possible that the very low thermal conductivity of
ZP131 material was improved by the application of the mould coating, and
that this produced an improved heat transfer performance for ZP131. Thus
the cooling rates of ZP131 and ZCast501 were similar, and the differences
were statistically insignificant.
41.81
42.26
43.20
41.60
42.00
42.40
42.80
43.20
ISOMOL ZIRCOAT MAGCOAT
S/N
rat
io
Mould coating
92
While the castings produced in the ZP131 and ZCast moulds were similar in
as-cast strength, the castings produced in the Silica foundry sand gave the
lowest UTS values in the case of AZ91 and A356. This result was related to
the mould design and dimensions used. The RP moulds were constructed to
the constraints of the 3D printer with a specially designed sprue and riser. To
construct a similar mould in traditional sand required a unique pattern to be
constructed for the moulds. Limited time and resources meant that a pattern
could not be made exactly to the dimensions of the 3D printed mould. The
larger mould thickness of the foundry sand mould resulted in lower thermal
gradients and metal flow velocities. This would produce a longer solidification
time, resulting in more grain growth. The grain size and Dendrite Arm Spacing
(DAS) is inversely related to the UTS and cooling rate in Mg and Al alloys [65,
67, 69, 70], thus meaning a lower UTS.
Figure 3.15 S/N ratio of the mould material factor levels in the UTS response
The actual tensile strength results from the Al and Mg castings produced in
the 3D printed moulds were similar in magnitude. Generally the strength
values in both RP moulds were very similar with all three alloys. The silica
foundry mould however produced lower tensile properties for AZ91 and A356
alloys, with improved values of SC1 to those obtained in the RP moulds. The
following is a general overview of the achieved strengths of the different alloys
used.
42.7542.71
41.8141.60
42.00
42.40
42.80
43.20
ZP131 ZCAST501 Silica Sand
S/N
rat
io
Mould material
93
Magnesium
Overall, the UTS of the cast Mg averages of AZ91 were of 156 MPa and 150
MPa achieved with AZ91 from ZP131 and ZCast moulds respectively. The
maximum values achieved with AZ91 were 170.63 and 150.87 MPa in the
ZP131 and ZCast moulds respectively. Castings produced in the silica sand
moulds were slightly lower with an average of 127 MPa with AZ91. The SC1
Mg grade produced consistent results, with 130 MPa and 127 MPa for
castings produced in ZP131 and ZCast moulds respectively. However, the
best results were achieved with the SC1 alloy in the silica sand mould with an
average as-cast strength of 157 MPa.
Typical values for sand cast AZ91 alloys obtained in traditional sand moulds
reveal that the minimum as-cast UTS should be around 160 MPa [27, 71].
Though the average UTS values achieved when casting into the ZP131 and
ZCast moulds were slightly lower, in some cases this was exceeded. From
this perspective, the RP moulds were capable of producing adequate casting
properties in terms of UTS. Furthermore an in-depth evaluation of AZ91 alloy
cast in traditional plaster moulds, showed that RP plaster moulds gave higher
UTS of the as cast samples. Work into the effects of grain refinement and
increased cooling rate led to higher strengths. However the AZ91 magnesium
castings produced in traditional plaster moulds were found to have an
average as-cast UTS of only 110MPa [36] compared to 156 MPa in ZP131
plaster moulds.
The SC1 Mg alloy was not able to be compared to values found in the
relevant literature directly, as no as-cast data could be soured. Typical values
found in literature [56, 72] however, were after the samples had undergone a
solution heat treatment and precipitation hardening. The heat treated SC1
samples found in the literature [56], have a 0.2% Proof stress of 130MPa with
a UTS of 202MPa. The samples prepared in the 3D printed moulds attained
an average UTS value of around 130MPa as-cast, with a maximum of
138MPa, 153MPa and 164.73MPa achieved in the ZP131, ZCast moulds and
Silica moulds respectively.
94
Aluminium
The tensile results were lower than those achieved with Mg alloys but were
close to the values achieved in traditional sand casting. Typical as-cast
strength of A356 alloy (Al-Si7%) was 130-150MPa [73] and castings produced
in the ZP131 moulds averaged 127MPa but did achieve a maximum of 141
MPa. The castings processed in the ZCast moulds improved on this with an
average of 132 MPa, and a maximum of 138 MPa. Unfortunately, the castings
produced in the Silica sand were poor and averaged only 93 MPa with a
maximum of 102 MPa. The large presence of hydrogen porosity present in the
cast samples meant that the microstructure was significantly weakened
leading to low UTS and elongation percent properties. This was perceived a
function of the melt treatment and casting setup, rather than that of the RP or
foundry sand moulds, with inefficient lance degassing unable to remove
dissolved hydrogen.
Tensile testing conducted from A356 tensile test parts, produced at the
supporting foundry, Centracast, attained 142 MPa. These underwent correct
melt treatment, which involved rotary degassing, reduced pressure testing on
the molten metal and the use of cleaning fluxes. This can be considered a
benchmark, as rotary degassing is considered the most efficient method of
reducing hydrogen from Al melts.
3.5.3 Percent Elongation
Table 3.7 shows the ANOVA conducted on the percent elongation values
from the tensile testing. This table shows that only the alloy factor was
significant, with a P value less than 0.001. This low P value results in a
confidence value at the 99% level. When considering the S/N ratio (Figure
3.16a) of the alloy factor, the SC1 Mg alloy achieved the maximum percent
elongation. This was followed closely by the AZ91 alloy, with the A356
aluminium alloy giving the lowest S/N ratio. Traditionally, AZ91 and A356
have close values of as-cast ductility [55]. In this case, the fact that A356 was
not effectively degassed, producing hydrogen embrittlement, could have
95
accounted for the variation obtained.
Table 3.7 ANOVA table for percent elongation response model
Rank Source DOF SS Variance FRatio P value Percent contribution
1 Alloy 2 180.84 90.42 101.39 0.00 92.42
2 Pool Temp 2 11.27 5.63
5.76
3 Pool Mould 2 1.58 0.79
0.81
4 Pool Coating 2 0.21 0.10
0.11
Pooled Error 6 1.78 0.89
Total 8 195.68
100.00
(a) (b)
(c) (d)
Figure 3.16 S/N ratios relating to the percent elongation values at factors (a) Alloy type
(b) Mould material (c) Mould coating and (d) pouring temperature
In comparison to the alloy factor, the overall variance accounted for by the
other factors was negligible. These factors were subsequently pooled as the
4.71
8.62
-2.22-4.00-2.000.002.004.006.008.00
10.00
AZ91HP SC1 A356
S/N
rat
io
Alloy type
4.13
3.13
3.84
3.00
3.20
3.40
3.60
3.80
4.00
4.20
ZP131 ZCAST501 Silica Sand
S/N
rat
io
Mould material
3.72
3.51
3.88
3.40
3.50
3.60
3.70
3.80
3.90
4.00
ISOMOL ZIRCOAT MAGCOAT
Axi
s Ti
tle
Mould coating
5.28
3.02
2.81
2.00
2.50
3.00
3.50
4.00
4.50
5.00
5.50
690°C 730°C 770°C
S/N
rat
io
Pour temperature
96
experimental error. In particular, mould coating and mould material both
contributed minimally to the overall variance. For completeness however, the
S/N ratios of these factors are shown above in Figure 3.16b-d. The highest
S/N ratio for each factor was the ZP131 for the mould material, Magcoat for
the mould coating and 690° for the pouring temperature.
In terms of percent elongation, interestingly SC1 outperformed AZ91 and the
A356 alloy. The HCP structure of Mg would tend to indicate a lower ductility
when compared to A356 and in practice A356 has been found to be more
ductile due to its Face Centred Cubic (FCC) crystal structure, especially after
a heat treatment. However, SC1 is a special alloy developed by the Australian
research institute CAST, to increase the creep properties of magnesium in
high temperature applications. The addition of special alloying elements such
as Zinc (Zn), Lanthanum (La), Cerium (Ce) and Zirconium (Zr) create a
completely different secondary phase which is intended to act as a more rigid
interlocking medium to resist grain movement [74]. This increased ductility
was most likely a result of this interlocking nature of the secondary phase.
Usual alloys of Mg with Al and zinc such as the AZ series alloys create the
well established secondary phase, Mg17Al12. This intermetallic created what
many researchers have described as a brittle and weak phase unable to
provide much resistance to grain movement when under load, especially at
elevated temperatures [75]. However, the addition of Al in Mg alloys has been
a requirement due to the Al increasing important casting properties such as
fluidity and corrosion resistance.
Temperature of pouring was also established as a significant variable with
respect to elongation percent. In addition to this it can be seen that in all the
other responses, the lowest casting temperature, 690°C, gave the highest
ranking level of the temperature variable. This is probably due to the reduced
solidification time when casting the metals at lower temperatures. This would
have resulted in less time for dendrite growth, and ultimately grain size, which
has a significant effect on the mechanical properties of Al and Mg casting
alloys [65-67, 70, 76]. Also, the solubility of hydrogen increases [77] when the
melt temperature is increased and this may have resulted in more porosity in
97
castings poured at higher temperatures.
3.5.4 Brinell Hardness
The variance analysis of the Brinell hardness presented in Table 3.8, showed
that the alloy factor accounted for a significant portion of the variance. This
significance was tested at the 95% confidence level. When examining the
individual levels of the alloys used, the A356 alloy had the highest hardness
value, which was followed by the AZ91 alloy grade.
Table 3.8 ANOVA of Brinell hardness response
Rank Source DOF SS Variance FRatio P value Percent contribution
1 Alloy 2 5.96 2.98 6.69 0.05 73.66
3 Mould 2 1.24 0.62 1.39 0.35 15.33
3 Pool Temp 2 0.27 0.13
3.30
4 Pool Coating 2 0.62 0.31
7.71
Pooled Error 4 0.89 0.45
11.01
Total 8 8.09
100.00
The mould material factor was the second ranking factor in the ANOVA table.
This factor had an influential effect and the S/N ratio showed that the
ZCast501 moulds produced the optimum hardness values. This was followed
by the ZP131 and Silica foundry sand moulds respectively.
98
Figure 3.17 S/N ratio of the alloy type factor levels in the Brinell hardness response.
The ANOVA in Table 3.8, shows that the effects of mould coating and pouring
temperature on the hardness were insignificant. The lowest pouring
temperature of 690°C produced its highest S/N ratio, but the contribution to
overall variation in hardness values was low. Disregarding the large influence
of the alloy factor in the ANOVA table, little variance was accounted for by the
remaining factors.
Figure 3.18 S/N ratio of the Mould material factor levels in the Brinell hardness
response.
35.27
33.65
35.47
32.50
33.00
33.50
34.00
34.50
35.00
35.50
36.00
AZ91HP SM1 A356S/
N r
atio
Alloy type
34.80
35.26
34.35
34.00
34.20
34.40
34.60
34.80
35.00
35.20
35.40
ZP131 ZCAST501 Silica Sand
S/N
rat
io
Mould material
99
(a)
(b)
Figure 3.19 S/N ratios relating to the Brinell hardness values at factors (a) Mould
coating and (b) pouring temperature.
The Brinell hardness response was actually an obvious result with the
predictably harder aluminium receiving some significance in terms of alloy
type. Ignoring the alloy type in the ANOVA table however, it can be seen that
there were no other factors that were even remotely significant and the
pouring temperature and mould coating variables were pooled as error terms
due to their relative size to the much larger alloy effect. The only other
variable having some effect on the hardness was the mould material, however
even this effect was only accounting for 5% of the total variation.
In terms of actual hardness values, the A356 alloy gave an average Brinell
hardness value of 59.62 with AZ91 a close second with 58.06. SC1 was the
34.43
35.03 34.94
34.1034.2034.3034.4034.5034.6034.7034.8034.9035.0035.10
ISOMOL ZIRCOAT MAGCOATS/
N r
atio
Mould coating
34.99
34.57
34.84
34.40
34.50
34.60
34.70
34.80
34.90
35.00
35.10
690°C 730°C 770°C
Sn r
atio
Pour temperature
100
softest alloy with only a BH value of 48.17. These values were in close
comparison to values obtained in literature. These were 60 for A356 and 50-
65 AZ91HP [55].
3.6 Metallographic Analysis
All cast samples were subject to metallographic analysis subsequent to
mechanical testing. The following sections describe the metallurgical and
metallographic aspects of the cast alloys. Attempts were made to compare
results with existing literature and to correlate the microstructures and
mechanical properties observed.
3.6.1 ASTM Grain Number
Grain counting was conducted using the relevant ASTM standard [61]. The
counting of grains quantified the number of grains present within a specific
area and with the use of reference charts the ASTM grain number can be
calculated. The grain number is the inverse of the grain size and allows the
average values of grain size and shape to be determined.
Figure 3.20 below shows the calculated ASTM grain numbers of the static
castings produced in the various moulds. The grain number was similar when
the three alloys were cast in the RP moulds. However, the castings produced
in the Silica foundry sand gave slightly higher values with respect to the two
Mg alloys. Interestingly, the Al castings gave similar grain numbers when
produced in each mould, although lower than for Mg alloys.
101
Figure 3.20 Average ASTM grain number produced in different moulds
The grain numbers for each alloy quantifies their microstructural granular size
and is directly related to the tensile properties of the castings. Microscopic
analysis of each alloy follows, utilising light microscopy and SEM.
3.6.2 Magnesium alloy: AZ91
The equilibrium phase diagram for Magnesium-Aluminium alloy systems is
shown in Fig. 3.17 and is a typical binary system with solid solutions,
displaying an intermetallic phase and an eutectic reaction. Photomicrographs
of Figure 3.22 the microstructures of different grades of magnesium alloys
obtained while casting into moulds under varying conditions as per the
Taguchi L9 design. In all cases the white areas represent the pro-eutectic α
Mg phase, and the darker areas represent the eutectic mixture of the α phase
and the intermetallic phase, Mg17Al12.
The cast structures produced by the ZP131 mould have the coarsest primary
grain structure, which is probably due to the poor heat transfer characteristics
of the plaster mould material. This slow cooling rate promotes a more
homogenous dispersion of the intermetallic phase, Mg17Al12. The Hexagonal
Close Packed (HCP) structure of the Mg crystals results in a limited number of
slip planes and resistance to fracture is primarily provided by the grain
0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
5.00
ZP131 ZCAST Silica
AST
M g
rain
siz
e
AZ91HP SC1 A356
102
boundary phase [75]. The intermetallic phase in AZ91 alloys is a Body
Centred Cubic (BCC) structure which is incompatible with the HCP Mg
structure. This difference results in the segregation of the intermetallic around
the α-Mg grains. It has been shown that the bond between the Mg/Mg17Al12
interface is fragile and the intermetallic itself is a relatively soft phase. This
results in a perfect location for cracks to propagate from, with the formation of
micro cracks commonly observed at the Mg/Mg17Al12 interface.
Intermetallic Phase in Al-Mg Alloys
The role of the intermetallic in the microstructure has a large effect on the
mechanical characteristics of the AZ91 alloy. Also, the relative volume and
morphology of the intermetallic phase influences the mechanical properties
[75]. The formation of the cast microstructure of AZ91 clarifies the role of the
intermetallic phase. Commercial Mg casting alloys typically contain a small
eutectic volume fraction of Al to avoid increasing the concentration of the
intermetallic phase described above [70]. With reference to the equilibrium
cooling diagram shown below in
Figure 3.21, the eutectic reaction should only occur when the Al content is
about 13 percent weight [70]. However, eutectic structures were observed in
alloys with an Al content much lower than this, and this is shown in the results
in Figure 3.22. The eutectic structures commonly seen in Mg-Al alloys are
known as partially or fully divorced eutectic structures. The difference
between the partially and fully divorced structures is that the partially divorced
eutectic structures tend to contain islands of the eutectic in the intermetallic
phase; whereas, the fully divorced structures surround the primary Mg
dendrites with the intermetallic phase existing as separate phase or layer
between the primary dendrites.
103
Figure 3.21 Magnesium-Aluminium binary phase diagram [78]
The larger volume fraction of the primary dendrites results in the eutectic
mixture solidifying in small inter-dendritic regions [70]. When the alloy
solidifies, small trapped pockets of the intermetallic phase exist, which require
only a small amount of undercooling for nucleation and growth. This is said to
be sufficient to cause the remaining liquid between the primary dendrites to
solidify. When a large number of pockets are formed by the dendritic
structure, each isolated pocket requires this nucleation for the intermetallic
phase. This causes partially and fully divorced eutectic structures to form as
the undercooling required is larger for nucleation than it is for growth.
Higher cooling rates also assist in the formation of more divorced structures
as dendrite structure becomes more „branched‟. This branched structure then
isolates the eutectic into smaller volumes, which require a higher amount of
undercooling for the nucleation of the intermetallic phase, which results in
divorced eutectic structures. The resulting divorced eutectic structures in slow
cooling castings (such the ZP131 material), results in a more homogeneous
dispersion of the intermetallic phase, leading to increased resistance to
deformation under load [70].
104
(a) X100 (b) X400
AZ91, ZP131 mould, at 690°C, Isomol 200 coating (σUTS = 156.4MPa δ = 2.17% HB =
56.8 Grain size: 1.11)
(c) X100 (d) X400
AZ91 casting produced in ZCast501 mould cast at 740°C. The mould was coated with
Magcoat. (σUTS = 150.87 MPa δ =1.51% HB = 60.53 Grain size: 1.29).
(e) (f)
AZ91 casting produced in Silica sand foundry mould, cast at 770°C. The mould coating
was Zircoat (σUTS = 127.1MPa δ = 1.54% HB = 56.82 Grain size: 3.09)
Figure 3.22 AZ91 microstructures in (a&b) ZP131 (c&d) ZCast and (e&f) Silica foundry
moulds
175μm 45μm
105
This may explain the increased tensile properties of the slower cooling
castings produced in the ZP131 moulds over those from the ZCast and silica
sand moulds.
Although the average grain number of the castings produced in the silica
foundry sand was considerably higher (smaller grains), this effect may be
negated by the isolated dispersion of the intermetallic phase. This non-
homogenous dispersion may have resulted in a weakened structure due to
lower resistance to grain boundary movement. This has also been promoted
[70] as a problem of divorced eutectic structures (higher cooling rates e.g.
silica sand) in structural applications because of a decrease in the ductility of
the alloy. This result was also seen in the above results, with a clear drop in
the elongation percent of the casting produced in the silica sand moulds.
Thus, it appears that to obtain a Mg casting with suitable tensile properties,
the intermetallic phase must be evenly dispersed to resist grain boundary
movement [36].
The photomicrographs below in Figure 3.23 show both partially and fully
divorced eutectic structures. It has been shown [79] that faster cooling rates
tend to form fully divorced eutectic structures in Mg-Al alloy systems. Castings
produced in RP mould resulted in partially divorced (low cooling rate)
structures whilst those produced in silica foundry moulds produced fully
divorced structures (high cooling rate).
106
(a) Partially divorced – slow cooling (plaster)
(b) Fully divorced - Fast cooling (silica sand)
Figure 3.23 Micrographs showing (a) partially divorced and (b) fully divorced eutectic
structures in AZ91 magnesium alloy
Overall Microstructure of Mg Alloys
The overall structure of the primary Mg dendrites revealed a globular, coarse
grain structure with no evidence of secondary arm growth in RP moulds.
Divorced
Massive
intermetallic
phase
45μm
45μm
107
Microporosity was also present in the microstructure of the cast AZ91.Figure
3.24 below shows suspected microporosity from Hydrogen present in the
atmosphere, resulting in small levels of hydrogen dissolving into the melt.
Upon solidification, the dissolved hydrogen is trapped, which forms micro-
voids in the cast structure.
Figure 3.24 Photomicrograph of AZ91 casting showing areas of micro-voids (X50
magnification)
The presence of this porosity has been linked to a decrease in the mechanical
properties of cast Mg alloys. The level of porosity (% of cross sectional area),
has a linear dependence with yield strength in permanent mould cast AZ91
alloys [80]. Non linear and power law relationships of porosity were also
established for UTS and % elongation of the tensile test specimens [67, 80].
Further quantification of the cast structure by SEM in this study revealed small
levels of porosity located on the fracture surface of the tensile test specimens
as shown below, in Figure 3.25. There was little evidence of the ductile dimple
structure found in plastically deformed fractured surfaces. The fractured
surfaces appear to have been caused by transgranular brittle fracture
(cleavage) [81]. Other observations from tensile testing, such a relatively flat
fractured surface, limited ductility and audible fracture confirmed the brittle
fracture mode. Literature [81] also supported this finding, with the fracture of
Mg alloys explained as being brittle in a cleavage or quasi-cleavage mode
300μm
108
due to limited ductility. The photomicrograph of the fractured surface obtained
by Zeng et al is reproduced in Figure 3.26a, which is similar to the current
results shown in Figure 3.26b.
Figure 3.25 SEM photographs showing porosity present on the fractured surfaces of
the AZ91 Castings (X130 (left) and X500 (right) magnification respectively).
.
(a) From current work (b) Zeng et al [81]
Figure 3.26 SEM fracture surface showing quasi-cleavage fracture surface in (a)
current work and (b) Literature
109
3.6.3 Magnesium alloy: AM-SC1
Unlike the AZ91 alloy system which utilises Al as the primary alloying addition
to the base Mg, SC1 comprises rare earth elements such as Nd, Ce, La, plus
Zn, Zr and Mn. Zn, Zr and La contain about 0.5% of the total weight, with
higher concentrations of Nd (1.7%) and Ce (0.7%). Al is not present and this
alloy system falls under the umbrella of Zr containing alloys. SC1 was
developed at the CAST Metals Institute with the objective of developing a high
temperature, creep resistant alloy capable of being cast into sand moulds for
automotive applications, primarily engine blocks and gearbox housings [82].
This alloy has an intermetallic phase which is located around the grain
boundaries, which reduces grain boundary sliding at higher temperatures.
The grain number of SC1 alloys was larger (smaller grains) in the silica sand
when compared to those used in RP moulds. The silica moulds produced both
increased the UTS and percent elongation. The absence of the Mg17Al12
intermetallic reaction, shown to be a relatively weak strengthening phase, is a
probable reason for this improvement. The RP moulds produced lower UTS
and percent elongation when both compared to the AZ91 alloy and to the SC1
castings produced in the Silica sand.
Overall Microstructure of SC1
SC1 alloys were observed to have globular primary dendrites, surrounded by
the Mg-RE intermetallic phase, as shown below in Figure 3.27a-f. The
castings produced in silica moulds showed consistently higher UTS and
elongation values compared to the RP moulds, being 157MPa and 3.35%
respectively. On the other hand, the castings showed in the ZCast and ZP131
moulds, showed consistently lower UTS and elongation values of 132 and
130MPa, and 2.28 and 2.56% respectively. The microstructure of castings
produced by 3D printed moulds was similar to traditional produced SC1
castings, as noted in the literature [56, 72], supporting the results of this work.
The literature samples showed typical α-Mg dendrites with a surrounding
intermetallic phase, chemically reported as Mg12La0.43Ce0.57 [56].
110
.
(a) (b)
SC1 magnesium alloy cast in ZP131 plaster moulds at 730°C. The mould was coated
with Zircoat (σUTS = 130.02 MPa δ = 2.56% 48.87 Grain size: 3.08)
(c) (d)
SC1 casting poured into ZCast501 moulds at 770°C. The mould was coated with Isomol
200 (σUTS = 128 MPa δ =2.28% HB = 48.16 Grain size: 3.21)
(e) (f)
SC1 casting produced in Silica foundry sand moulds at 690°C. The mould was coated
was Magcoat (σUTS = 157.38MPa δ = 3.35 HB = 47.47 Grain size: 4.6).
Figure 3.27 Photomicrographs of SC1 produced in (a&b) in ZP131 and (c&d) ZCast501
and (e&f) Silica foundry sand moulds
175μm 45μm
111
Overall, the SC1 grain numbers were higher (ASTM E112), than those of
AZ91, due to the grain refining properties of Zn. However, this only translated
into increased tensile properties compared to AZ91 using silica sand moulds
and not the RP moulds. This suggests that to achieve equivalent tensile
properties to AZ91 with SC1 a significant increase in the cooling rate must be
achieved (even smaller grain size). This reversal in performance using SC1
as against AZ91 may also be due to additional RE intermetallic phase
contributions in SC1.
Subsequent SEM analysis suggests transgranular failure as the mechanism
of fracture of this magnesium alloy. SEM photographs clearly show fracture
surfaces that are consistent with transgranular failure, as shown in Figure
3.28. Similar to AZ91, small microporosity can be observed on the fracture
surface.
(a) (b)
Figure 3.28 SEM photographs showing fracture surface on the primary (a) Mg dendrites
and (b) more ductile regions on the same sample (magnifications are X500 and X4000
respectively).
3.6.4 Aluminium: A356
The photomicrographs obtained with A356 cast under different conditions are
shown below, in Figure 3.29a-e. Inspection of the microstructure showed
unusual plate like Silicon (Si) eutectic morphologies, coarse grain structures
and large porosities in the Al-7%Si castings. Also, the photomicrographs
112
showed the presence of Iron (Fe) intermetallics in the cast alloys. Subsequent
SEM photographs of the fractured surface showed large pores. The presence
of coarse Si eutectic particles, iron intermetallics and porosity in cast Al alloys
all have adverse effects on the material quality and strength [57, 65, 69, 76,
83, 84].
These properties of A356 must be placed in the context of Al casting alloys.
This alloy is a widely used hypoeutectic alloy, which includes primary Si
around 7% weight and 0.03% Strontium (Sr). The microstructure of this alloy
consists of Al dendrites, and an Al-Si eutectic. A356 has a modified eutectic
structure because of Sr content. An unmodified eutectic structure results in
silicon particles which have a plate like morphology.
The modification of these plate-like Si particles to rounded particles results in
improved strength and elongation values. Permanent mould casting
demonstrates the influence of the eutectic modification [83] on the mechanical
properties of Al-7%Si alloy. It was shown that in unmodified castings the UTS
was 77MPa, whereas partially and fully modified samples were 147 and 188
MPa respectively. The eutectic modification results in a „composite like‟
structure which increases UTS, elongation, hardness and machinability [83].
Other sources [85] have identified that the mechanical properties of Al-Si
castings are dependent on the size, form and distribution of the silicon
particles, together with the level of porosity, Dendrite Arm Spacing (DAS) and
eutectic morphology.
The morphology of the plate like constituent Si particles appears to be a result
of no Sr modification. This may have occurred because of strontium burn off
which is a known phenomenon. The result was large brittle plate like Si
particles which were detrimental to the mechanical properties of the cast parts
[85]. The shape and size of these Si particles is important as increasing DAS
has been linked to increasing size of the Si particles [86]. Modification of the
Si particles was found to decrease the overall Si particle size and number of
particles per mm2 [86].
113
(a) (b)
A356 Aluminium castings produced in ZP131 plaster moulds, poured at 730°C. The
mould coating used was Zircoat (σUTS = 140.23MPa δ = 0.96% HB = 60.53 Grain size:
2.45).
(c) (d)
A356 castings produced in ZCast moulds poured at 770°C. The mould coating was
Isomol 200 (σUTS = 134.30MPa δ = 0.74% HB = 66.70 Grain size: 2.45).
(e) (f)
A356 castings produced in Silica foundry sand moulds, poured at 690°C. The mould
coating used was Magcoat (σUTS = 102.26MPa δ = 0.89% HB = 52.65 Grain size: 2.18)
Figure 3.29 Microphotographs showing A356 cast structures in (a&b) ZP131 (b&c)
ZCast501 and (d&e) Silica foundry sand
45μm 175μm
114
Eutectic morphology is also important inasmuch as shrinkage and porosity are
adversely affected by the eutectic solidification. This proceeds through
solidification of primary Al dendrites until the dendrites impinge on each other
[85]. At this point, dendritic mobility is reduced. Interdendritic feeding the
primary mechanism by which the contraction of the metal is compensated for,
as the volume reduces upon freezing. This feeding results in the flow of
eutectic fluid, which indicates that the eutectic formation and growth are
important to the fluid flow and ultimately the cast mechanical properties.
(a) Typical modified eutectic structure found from literature [60].
(b) Static A356 casting showing sharp silicone eutectic (X200)
Figure 3.30 Al-Si alloy, showing (a) fine modified eutectic silicon and (b) coarse
eutectic silicon.
45μm
115
The plate morphology is detrimental to the cast alloy strength. Proper
modification is shown above in Figure 3.30a, with the unmodified structures,
from the current work, shown in Figure 3.30b. This mechanism has been
explained by the reduction of stress concentrations present at the sharp Si
needle particles. The presence of these stress concentrations reduces
material toughness, leading to typical brittle fracture [84]. Further, it was found
[84] that in the unmodified alloy, the damaged Si particles were present on the
fracture surface of tensile tested specimens, indicating that the cracks which
caused fracture initiated from these Si particles. It was concluded that in
unmodified alloys the Si particles have a plate like structure, from which
fracture is likely to initiate and propagate. Conversely it was determined that
modified alloys have a ductile fracture mode, with considerable plastic
deformation seen before fracture, and that the modified eutectic structures
indeed have a strong relationship with fracture behaviour [84]. In this work,
all Al alloys were seen to have a transgranular fracture mechanism, with the
unmodified alloys producing intergranular fracture modes. These results
indicate further that the eutectic structures are weak and are a typical source
of fracture, as cracks propagates around the dendrites in unmodified alloys.
Proper modification of the eutectic has also been found to not only refine the
eutectic Si particles but also the Iron plate like particles, which reduces
embrittlement of the cast specimens [87]. Further research [88] showed that
the size and shape of the eutectic silicon particles and Fe intermetallics affect
the tensile and fracture properties of A356 alloys. In unmodified alloys (heat
treated, T6), the eutectic silicon particle size and aspect ratio led to a
decreased ductility, accompanied by increased SDAS. At larger SDAS in
particular, ductility was low, which was synonymous with large elongated
silicon particles. The tensile properties were described as very dependent on
the SDAS, eutectic particle size and the aspect ratio of eutectic particles [88].
The presence of iron can also be manifested as large plate like iron particles
and these are similarly harmful to the cast strength [89]. The size of Fe
intermetallics in A356 aluminium alloys depends on the cooling rate and the
amount of iron present in the alloy. Cooling rates decrease the intermetallic
116
plate size and increasing iron content increases the plate size. Inverse
logarithmic relationships of Fe intermetallic particle length to percent
elongation and UTS were found fro of unmodified A356 permanent mould cast
samples.
Figure 3.31 Photomicrograph showing plate like Fe intermetallics in A356 casting
produced in 3D printed moulds
These Fe intermetallics were identified in the current work as β-Al5FeSi which
are detrimental to the mechanical properties of the castings and in particular,
particles over 70μm in length in A356 castings [89]. From the
photomicrograph obtained in the current study, and shown in Figure 3.31 an
identified Fe intermetallic particle of 200μm can be seen. Fe intermetallics
decrease ductility and tensile strength by „participating‟ directly when fracture
occurs [90]. Unmodified alloys are easily fractured under loads relative to the
Al and Si phases of the alloy. This is caused by the formation of micro-cracks
which provided a low resistance path for larger cracks to propagate through
[90].
As well as plate like Si and Fe in porosity also affects the mechanical
properties. The rounded pores observed in Figure 3.32 below, are due to the
presence of dissolved hydrogen, which enters the molten Al as the alloy
changes to the liquid phase. This change of phase dramatically increases
Fe intermetallic
200μm
117
hydrogen solubility. Upon solidification, impinging dendritic networks trap the
gas, leading to the formation of pores in the cast specimens.
a) Microporosity (b) Macroporosity
Figure 3.32 Un-degassed showing typical hydrogen porosity in (a) microstructure and
(b) macrostructure produced in RP moulds
The solubility of hydrogen is dependent on the partial pressure of the
hydrogen in the atmosphere. This is shown below in equation (3.3).
2HH S P
(3.3)
Where:
H = the amount of hydrogen in liquid Aluminium, cm3/100g
2HP = the partial pressure of H2 of gas in the atmosphere
Figure 3.32a&b above, show large porosities in the un-degassed specimens.
These samples un-degassed were examined to confirm that the porosity
present was indeed due to hydrogen. This is required because two types of
porosity form during solidification, namely shrinkage porosity and gas porosity
[68]. Evidence from literature on traditional sand casting shows that hydrogen
porosity formed is dependent on the solidification time as interdendritic
porosity is suppressed by rapid cooling, which impairs the transfer of
hydrogen into the liquid [64]. Slow cooling of the metal thus leads to
interdendritically distributed voids containing hydrogen.
300μm
118
It is difficult to distinguish between shrinkage porosity and hydrogen porosity
but hydrogen is often said to be „additive‟ to existing shrinkage porosity. This
is because the shrinkage void created upon solidification is caused by a drop
in pressure and this results in a reduction in hydrogen solubility in the
surrounding liquid which facilitates the precipitation of the hydrogen into the
forming void [64]. From experience and the literature hydrogen voids are
characterised by a round smooth surface in the defect void. Also the wide
distribution of the hydrogen voids provides much of the evidence as shrinkage
voids are more localised and not distributed evenly over the section being
considered. The voids shown in Figure 3.32a&b were evenly distributed
porosity, which is typically round in shape. This indicates that the porosity
present was hydrogen driven and discussions with foundry engineers
confirmed that these voids were indeed of the hydrogen type. The presence of
hydrogen and subsequent porosity in the cast specimens, led to reduced
tensile strength.
Literature reveals that porosity indeed produces „large decreases‟ in both
tensile strength and ductility [91]. Also, not surprisingly, porosity has a
detrimental effect on the fatigue life of Al-Si cast alloys, with regions of high
stress created around the pores present [92]. Other research [89] also
observed inverse logarithmic relationships of porosity pore size (mm2) with
elongation and UTS values from cast tensile test bars. Similar Al alloys such
as A356.2 (unmodified) were degassed and cast into permanent moulds and
the presence of pores in all cases led to decreased mechanical properties.
Large decreases in elongation and UTS were seen in samples with pores
over 50μm in length [89].
Subsequent SEM work on the static castings showed large porosities in the Al
castings obtained in the current experimental work. Typical porosities present
in a SEM photograph are shown below in Figure 3.33
Figure 3.33. The pore size is about 0.30mm (longest diameter). Literature
suggests that the overall porosity (represented as a percent of the fracture
area) of the cast A356 samples have been found to have a negative linear
relationship with UTS and an inverse parabolic relationship with elongation
119
percent.
Figure 3.33 SEM photographs showing large pores from su8spected hydrogen porosity
SEM analysis of fractured Al specimens found that the fractured surfaces
contained intergranular dimple type rupture. The dimples themselves were not
observed to be deep and exhibited only a partial rim around each dimple. This
indicated that the fracture had progressed in a specific direction. Conventional
fracture mechanics shows that typically, materials fail due to crack
propagation in the presence of local stress raisers in the material. In this case,
casting defects such as hydrogen porosity, entrapped oxides and
intermetallics, act as locations of high stress concentrations, from which
micro-voids can form and cracks can propagate.
The morphologies of the fractured surfaces are also influenced by the loading
type also. Fractured surfaces found from uniaxial tensile for example, results
in equiaxed dimples bounded by a lip or rim [93]. The SEM photos shown
below in Figure 3.34 depict this however, the rim does not extend all the way
around each dimple.
120
(a) (b)
Figure 3.34 SEM fracture surfaces showing (a) typical dimpled, ductile fracture surface
[93] and (b) a sample from the current research , exhibiting a lack of a rim around each
dimple.
Elongated as opposed to round dimples with an open end indicated the failure
mode was a type I „tear‟ fracture. This means that the crack originated at a
defect near the surface or from porosity present at the surface, allowing the
tearing action to propagate through the material.
Figure 3.35 Different fracture modes, with arrows showing the direction of the applied
loads [93]
Evidence of a crack path originating from porosity is shown at the bottom of
the pore. Thus the SEM evidence shows that, the examined fracture surfaces
appeared to be typical of brittle fracture. The specimen shown in Figure 3.36
did indeed have limited ductility with 0.74% strain measured from the tensile
test and the SEM photograph shows evidence of a pore intercepted by the
fracture surface. These shallow dimples were consistent with brittle fracture
initiated at pores present in the material. Large pores, shown in Figure 3.36
121
below are likely candidates for stress raisers and crack initiation points. The
absence of shear lips around the edge of the broken specimens also indicates
a general lack of ductility [93].
Figure 3.36 A Crack path seen to initiate from pores due to suspected hydrogen
porosity.
The combination of unmodified eutectic silicon, porosity and large grain
structure, resulted in brittle fracture in the Al castings produced by RP moulds.
This was a function of the melt quality rather than the RP moulds used.
3.7 Summary
Static casting of light metals, namely Aluminium and Magnesium was
successfully performed in 3D printed mould materials. Significant process
factors were linked mechanical properties and possible reasons for their
importance was conjectured. Generally, the cast mechanical properties were
comparable to values found from traditional sand casting, leading to the
conclusion that these RP moulds are suitable for the static processing of light
metals
122
CHAPTER 4 CENTRIFGUAL CASTING IN RP MOULDS
4.1 Centrifugal Casting
The Centrifugal Casting (CC) process is attractive because it is able to
improve cast properties by filling the mould with metal under pressure, due to
the absence of the centripetal force. This process also facilitates the filling up
of more complex and thin sections which may not be achieved by gravity
alone. The use of CC offers several mechanical advantages, such as
reduction of porosity, increased solidification rate, refined grain structure and
increased cast strength. Thus there are both potential design and economic
benefits with this casting process. Use of CC can also reduce the amount of
scrap in symmetrical components as removal of scrap is often more straight
forward, because there is only one central sprue and often only one gate to be
attached to each cavity. This reduces both material and labour costs,
increasing the overall casting yield. Further, the CC process can be utilised in
large scale production, including creation of automated casting processes.
Machines and operations can be numerically controlled and casting and
mould removal become mechanical actions. Castings can be repeated with
fine control over speed of rotation of the mould, pouring rates, cooling rates
and casting removal.
Apart from the increased pressures generated in CC, the casting process also
facilitates functional grading of materials. For example, in Aluminium-Silicon
Carbide (Al-SiC) alloys, hard and heavy SiC particles present in the alloy are
displaced at a greater rate than the lighter Al particles, leading to a greater
concentration of the hard SiC phase at the periphery of the casting with the
lighter more ductile aluminium phase located away from the outer surface.
The production of car wheels is a common example where a hard surface is
desired for the outer rim of the wheel to reduce wear associated with
operation, whilst a ductile region is desired at the centre to avoid brittle
fracture.
123
The ability of the 3D printing process to produce complex shapes has been
noted. The advantages of the CC process offer the opportunity of creating
complex strong castings in rapidly produced moulds. Also poor thermal
properties of the printed moulds may be improved by use of the CC process.
It is currently unknown whether in fact 3D printed moulds can be used in the
CC process, and this chapter presents work to experimentally investigate the
application of CC in RP moulds.
4.2 Background
The CC process is casting process whereby molten metal is poured in a
central sprue in the mould cavity, which is then filled under pressure, due to
the absence of the centripetal force. Filling under pressure creates castings
with superior mechanical properties and increased accuracy when compared
to traditional static sand casting [94].
Figure 4.1 Vertical semi-centrifugal casting setup [95]
The process consists of three primary techniques, these being [95]:
True centrifugal
Semi-centrifugal
Centrifuge casting
124
In true CC, cast iron and steel pipes, liners and sleeves were produced [95].
This can proceed around a vertical axis, as shown below in Figure 4.1, or
around a horizontal axis. The metal is cast with no internal core and the
molten metal solidifies in the shape of a tube. Semi-centrifugal casting usually
comprises a centrally placed cavity which is supplied by a central sprue. The
cast objects are typically small or medium sized castings. Finally, centrifuging
is when multiple cavities are placed a specific certain distance away from the
centre of the mould. The in-gates are connected to the central sprue, which
supplies molten metal to the cavity under pressure.
Figure 4.2 Illustration of the centrifuge technique with multiple castings placed around
the central sprue [95]
These CC techniques can utilise sacrificial moulds such as traditional green
sand or chemically bonded sand moulds. Other moulds used are plaster and
vulcanised silicone rubber moulds [94]. The use of sacrificial moulds in CC
may be applicable to creation of sand cast components comparable to those
produced by high pressure die castings.
125
4.3 Analytical Model
The initial approach to the CC process is through the development of a
mathematical model, to further understand the process on a more
fundamental level. Specifically, the relationship between metal velocity and
radial position with respect to pressure, is sought in a scientific manner, along
with the quantification of the magnitudes of the resulting metal velocities and
pressures.
The basis of using the CC technique is centred on the ability to fill a mould
under pressure, resulting in improved mechanical properties. This pressure is
created by increasing the fluid velocity inside the rotating mould, caused by
the absence of the centripetal force (informally known as the centrifugal
force). The increased fluid pressure developed, allows thin and complex
sections to be filled, as the pressure present in the fluid overcomes the
surface tension of the molten metal. This is advantageous to static casting,
which utilises only the force of gravity, requiring a large metallostatic head to
overcome the surface tension, to force the metal into thin sections.
Although the basic understanding of increasing rotational speed results in
higher fluid velocities, the actual magnitudes of these velocities and pressures
require investigation, into not only their quantity, but the nature of their
mathematical relationship to increasing rotational speed. To gain further
clarity and understanding, with respect to the current casting setup, a second
order differential equation was developed to predict the magnitude of these
properties at a given radius. This allowed the calculation of the pressure of the
metal at given radius as a function of time. These values were then cross-
referenced to the mechanical properties of the castings gained from the
experimental testing, in the attempt to link improved mechanical properties to
increases in rotational speed (increased filling pressures). Further, the effect
of rotational speed on the microstructural features of the cast parts, such as
defects and grain size, was of importance to help understand the influences of
the CC process on the properties of light metal alloys (section 4.7.1). Lastly,
the ability of these moulds to withstand the calculated pressures can then be
formulated and their suitability assessed, in the context of CC of light metal
126
alloys. The resulting range of influential rotational speeds from the
experimental results could then represented as a pressure and possibly used
in future work. However, due to restrictions with the respect to the 3D printer
build size and maximum rotational speed of the CC machine, speeds ranging
from 150-450RPM were analysed with at various radii (up to 90mm) from the
centre of rotation.
To quantify rotational speed, a particle with unit mass located at a fixed radius
from the centre of rotation was considered. This particle was confined within a
fixed channel which acted as the cylindrical portion of the rotating mould.
Figure 4.3 below shows the schematic of the experimental model, with the
free body diagram of the particle and the forces acting on an element of liquid
during the CC process. The model assumes that the fluid is not compressible,
and considers only the action of the centrifugal force. The element of liquid
shown is constrained in a radial channel as designated in the mould design.
Once spinning, the particle experiences acceleration due to the absence of
centripetal force. This has the effect of moving the fluid radially outward for,
the centre along the channel.
Figure 4.3 Schematic of the forces acting on an element of liquid under the absence of
centripetal forces, at constant angular velocity
Initially, neglecting friction experienced at the mould wall, the following
differential equation applies:
127
From Newton‟s second law of motion:
2
x 2
d rF =m.
dt (4.1)
Assuming the radial direction as the x direction, the equation (4.1) becomes:
22
2
d rr
dt (4.2)
Rearranging yields
22
2
d rr 0
dt (4.3)
Equation (4.3) is a homogeneous, second order, ordinary differential equation,
which has the following characteristic form:
2 2 0 (4.4)
Rearranging equation (4.4) gives:
2 2= (4.5)
The roots of equation (4.5) are then = ±
From this the general solution shown in equation (4.7) for this differential
equation is used to develop the expressions for radial position, velocity and
acceleration as a function of time.
1 2x t x t
1 2y(x) = C e +C e (4.6)
Substituting radial position (r) for y, time (t) for x and w for λ yields the general
solution.
wt - wt
1 2r(t) = C e +C e (general solution) (4.7)
128
To determine the constants C1 and C2, initial and boundary conditions are
implemented. From the initial condition, it is known that at time = 0, the radial
position equals 0, thus:
1 2r(0) = C + C (4.8)
Rearranging gives:
1 2C = -C (4.9)
Differentiating the equation (4.7), the velocity relationship is determined.
wt - wt
1 2
dr= C we - C we
dt (velocity expression) (4.10)
Applying the second initial condition, it is known that the initial entry speed of
the molten metal is V0. This can be determined from the falling height from the
crucible to the entry of the mould (sprue). Knowing this value, and that it
occurs at time equal to zero allows the constants C1 and C2 to be found.
Thus:
0t=0
dr= V
dt Giving
0 1 2V = C w- C w (4.11)
Rearranging gives 0 1 2V = C w- C w thus, (4.12)
01 2
VC = + C
w (4.13)
Substituting equation (4.13) into equation (4.9) yields:
02 2
V+ C = -C
w (4.14)
129
Rearranging equation (4.14) gives:
02
- VC =
2 w (4.15)
thus 01
VC =
2w. (4.16)
Substituting equations (4.15) and (4.16) into the (4.7) and (4.10) yields the
final expressions for radial position and velocity of the fluid element:
wt - wt
0 0V e V er(t) = -
2w 2w (radial position) (4.17)
And
wt - wt
0 0V e V edr= +
dt 2 2 (fluid velocity) (4.18)
From here the acceleration expression is obtained by differentiation equation
(4.18) giving:
wt - wt2
0 0
2
wV e wV ed r= -
dt 2 2 (fluid acceleration) (4.19)
With these relationships were established, the analysis of the different
velocities and pressures of flow can be determined at different radial positions
away from the central inlet. Knowing the values of these radial positions, the
time at these locations were solved iteratively in Microsoft Excel by the radial
position formula (equation (4.17)). The time found from this relationship is
then used in the velocity and acceleration expressions to find both these
values at the chosen radius.
However, before these values are obtained, the frictional resistance of the
flowing fluid over the mould surface has to be factored into the model. As the
overall aim of this model is to determine the pressure setup in the fluid at the
radius considered, the loss of pressure can be determined from Darcy‟s
130
equation, presented as a pressure head.
2
f
fLVh =
2Dg (Darcy‟s equation) (4.20)
Where:
f Friction factor
L = Pipe length (m)
V = Fluid velocity (m/s)
D = Pipe inside diameter (m)
g = Gravitational acceleration (m/s2)
Here the pipe length is the runner length into the cylindrical cavity and D is the
runner diameter. The friction factor is found from a semi-empirical equation
which is dependent on the mould surface roughness and the Reynolds
number. This equation is known as the Swamee and Jain equation [96] which
can be used for Reynolds numbers ranging from 5000-108. In the worse case
for velocity (no friction), the velocity can be computed from the analytical
model. From here the Reynolds number is then computed with the following
equation:
rVLRe =
m (Reynold‟s number) (4.21)
Where:
Re Reynolds‟s number
Density (kg/m3)
V = Velocity (m/s)
L = Pipe length (m)
μ = Dynamic viscosity (Pa.s)
The friction factor was then computed with a surface roughness value of the
mould material of 100μm. The Reynold‟s numbers found were in the region of
131
350000, which constituted turbulent flow. The following equation was used to
determine the friction factor.
x
0.9
e1 21.25=1.14 - 2log +
D Ref (4.22)
Rearranging gives;
2
x0.9
1f =
e 21.251.14 - 2log +
D Re (friction factor) (4.23)
Where :
f = Friction factor
x = Mould surface roughness (m)
D = Runner diameter (m)
Re = Reynolds number
Lastly, with the friction factor determined, the pressure loss due to the
frictional resistance of the metal flowing in the runner bar can be determined
by using the Darcy-Weisbach equation shown below. This yields a pressure
loss in terms of a head of the corresponding fluid.
2
f
fLVh =
2Dg (4.24)
Also Loss fP = gh
Thus 2
Loss
fLVP
2D (4.25)
This pressure loss is subtracted from the pressure created from the velocity
found using the velocity expression. This gives the final overall expression for
132
the pressure created at the end of the runner bar.
x x xTotal Model LostP P P (4.26)
Where:
xTotalP = Overall pressure created due to CC effect (KPa) (at a radius „x‟ away
from centre of rotation).
xModelP = Pressure from velocity determined from velocity expression (KPa)
xLostP = Pressure lost due to friction (KPa)
The pressure developed by the rotation of the mould is then computed as a
change in momentum from the entry point to the section under consideration.
The following equation relates force to momentum:
x xF m V (4.27)
And since
FP
A (4.28)
Therefore xx
mVP
A (4.29)
Where: Px = Pressure at any radius „x‟ from centre of rotation (Pa)
m = Mass of fluid (kg)
Vx = Velocity at radius x (m/s)
A = Cross-sectional area of duct (m2)
The overall pressure developed at any point x was shown to be:
total x LostP P P (4.30)
133
The initial velocity was evaluated to determine the actual velocity of the liquid
after undergoing the 90° turn into the runner. This was found from the initial
mass flow rate of the metal, which was assumed to be conserved. The initial
mass flow rate was set as 0.150 Kg‟s/s and the vertical distance from the top
of the pouring system to the mould cavity surface was 333mm (Figure 4.24).
The velocity of the fluid after undergoing this vertical drop was calculated with
the following equations:
The velocity at the cavity initial was found from:
V 2gH 2 9.81 0.333 2.55 m/s (4.31)
Where:
V = Melt velocity (m/s)
g = Gravitational acceleration (m/s2)
H = vertical fall height (m)
However, this is not the actual velocity entering the runner as compensation
for the 90° turn into the runner needs to be factored in. This results in the
reduction of the velocity to some value lower than 2.55 m/s, as determined
above.
To determine the true velocity, the equivalent friction of the bend in the pipe is
accounted for in the PLost equation. The Plost equation now takes the form of:
2
f
fLKVh
2Dg (4.32)
Where:
K = The Equivalent loss due to a fitting
134
And
LK f
D (4.33)
Where:
f = Fiction factor of pipe
L/D = Equivalent ratio from a pipe fitting, in this case a 90° bend.
It was found that the L/D for a 90° bend was 30. Also calculations made for
the friction factor was 0.033. Substituting these values into the Equivalent loss
equation, K equals 0.99.
Now the Pressure head loss due to this bend is computed, giving:
2 2
f
fLKV 0.033 0.03 2.55h 0.0205
2Dg 2 0.016 9.81 m
Converting this to a pressure:
LostP 2680 9.81 0.0205 538.97 Pa
The initial pressure resulting from the drop is calculated from the force acting
over the pipe area, since the velocity due to the initial fall is known and the
mass flow rate is also known, thus:
Initial 2
F Vm 2.55 0.150P 1902
A 0.000201D
4
Pa
Thus the overall Pressure allowing for the pipe bend is:
Total Initial LostP P P 1902 538 1364 Pa
Now this result can be used to establish the force, i.e.:
135
F PA 1364 0.000201 0.274 N
And since we know F mV :
F 0.274V 1.83
0.150m
m/s
Thus the initial velocity of liquid metal entering the runner is 1.83 m/s,
accounting for the pipe bend.
4.4 Results from the Analytical Model
The analytical model provided mathematical insight into the effect of rotating
the mould on the velocity and pressures developed in the liquid metal at
different stages of filling. The model was used to predict the pressures
present at the runner/cavity interface, which was situated 30mm from the axis
of rotation. The reason for predicting the pressure at this junction was firstly,
to obtain the magnitude of the pressure and secondly, analyse the effect of
the cavity angle and the runner/cavity ratio, with respect to the mechanical
properties, independently (as these effects were considered as separate
variables, as they occurred after this 30mm distance).
The below results showed that at a given radius, increasing the rotational
speed resulted in a velocity increase of the liquid metal and subsequently a
rise in the metal pressure in the cavity. These pressures were represented as
static pressure heads of the metal with the maximum pressure found to be
around 300mm Aluminium, which equates to 7.88KPa.
136
(a) (b)
Figure 4.4 Results of the analytical model showing (a) the velocity at different radii
from the centre of rotation and (b) fluid pressures developed at the same radii.
The model shows that the link between rotational speed and the melt velocity
(as function of time), is exponential in nature (equation 4.18), and provides the
basis for estimating the filling pressure. Also, increasing the radial position
away from the centre of rotation has the effect of increasing the amplitude of
the velocity (and thus pressure). The model thus indicates that to obtain
higher pressures, moulds should have increased radii to fully take advantage
of the CC process.
The experimental constraints present in this CC setup have been noted,
however this analytical model may be further used as a tool in future mould
design work, to ensure that correct rotational speeds are chosen. For
instance, typical speed ranges used in CC are represented as a ratio of the
centrifugal force to the gravitational force. This ratio yields the „G factor” which
is used as a general guideline for choosing the rotational speed. Values of 15-
50G have been quoted [53] in the relevant literature for semi-centrifugal
casting in sand moulds. Table 4.1 below lists the G values obtained in this
experiment at different speeds and radii from the centre of rotation.
0
1
2
3
4
5
0.01 0.03 0.05 0.07 0.09
Vel
oci
ty (
m/s
)
Radius (m)
150 RPM 300 RPM 450 RPM
050
100150200250300350
0.01 0.03 0.05 0.07 0.09
Pre
ssu
re H
ea
d (
mm
Al)
Radius (m)
150 RPM 300 RPM 450 RPM
137
Table 4.1 Values of G developed at different speeds with respect to different radii
Radius (m) 150 RPM 300 RPM 450 RPM
0.03 0.75 3.02 6.79
0.05 1.26 5.03 11.32
0.07 1.76 7.04 15.85
0.09 2.26 9.05 20.37
From the above table it is clear that only at a rotational speed of 450 RPM and
a radius of 0.07m or greater, was the minimum of 15G achieved in the current
setup. However, the mould created in this experiment was subject to the
constraints of the 3D printer build size and material cost. Typically much
larger moulds are used in actual production with radii up to 250mm [94]. The
analytical model is thus useful for calculating the predicted values of
pressures and values of G at specific radii from the centre of rotation. This
model however, constitutes only preliminary work to establish the linkage
between radial position, rotational speed and resulting velocities and
pressures in the mould. Other, more complex factors, such as density change
(compressibility) due to cooling of the flowing metal and the formation of
turbulence due were not considered, but these factors will have effects on
fluid flow.
4.5 Centrifugal Casting Experimental Design and Setup
The above analytical model has quantified a range of pressures due to the
effect of increasing rotational speed and radial position. To further assess the
effects of these pressures on the mechanical properties of the castings, the
CC trials were conducted, making use of analytical model and findings from
the static casting trials discussed in Chapter 3. This involved the use of ZP131
moulds in a statistical design of experiments known as a Box-Behnken
design, which uses three factors at three levels, creating a second order
polynomial expression in the three factors. This design was chosen over
Central Composite Design (CDD) used in Chapter 2 as only 15 trials are
required, as against 20 trial required for the CCD. While a number of factors
138
could have been considered for the CC trials, only three critical factors were
finally chosen, due to time and practical constraints. These factors are
presented below:
Speed of Rotation
The speed of mould rotation dictates the pressure generated in the flowing
metal. Faster rotational speeds thus result in greater pressures, which
influence solidification rate and improve the cast part density [44, 94, 97].
Variation in the rotational speed is of great interest and in these experiments
is the primary factor of importance for its impact on mechanical properties.
Cavity Angle
The direction at which the cavity angle is orientated with respect to the inlet is
also of importance, as it has a direct effect on how the metal enters the mould
cavity. There will be varying angles at which liquid metal enters different parts
of the casting, and variation in the angle of entry of the liquid metal into the
cavity may influence cast part properties.
Runner to Cavity (r/c) Ratio
Lastly, the ratio of the cylindrical runner diameter to the size of the final entry
section at the mould cavity represents the intensity of chocking, and may
result in changes in the properties of the cast parts. Relevant literature [94]
mentions the importance of ensuring directional solidification in centrifugal
casting, and, again variation in this factor may influence changes in the cast
part quality.
The experimental work involved simultaneous variation of the three factors at
three different levels, as shown in Table 4.2. The experimental design in
Table 4.3 shows both the coded variables and natural variables. Much like the
central composite design, this design also allows calculation of constants
shown in equation (4.34).
139
Table 4.2 Summarised factor and level combinations for the centrifugal casting trials
Factor Low Medium High
Mould rotational speed
(RPM)
150 300 450
Mould cavity angle (°) 45 90 180
Runner/cavity ratio 0.8:1 1:1 1.2:1
Table 4.3 Centrifugal casting experimental design table
Run X1 RPM X1 Angle (°) X1 Ratio
1 -1 150 -1 45 0 1:1
2 1 350 -1 45 0 1:1
3 -1 150 1 180 0 1:1
4 1 350 1 180 0 1:1
5 -1 150 0 90 -1 0.8:1
6 1 350 0 90 -1 0.8:1
7 -1 150 0 90 1 1.2:1
8 1 350 0 90 1 1.2:1
9 0 250 -1 45 -1 0.8:1
10 0 250 1 180 -1 0.8:1
11 0 250 -1 45 1 1.2:1
12 0 250 1 180 1 1.2:1
13 0 250 0 90 0 1:1
14 0 250 0 90 0 1:1
15 0 250 0 90 0 1:1
The Box-Behnken experimental design creates response surfaces for
quadratic models. The following polynomial expression is procedural
methodology for establishing the responses.
2 2
0 1 1 2 2 3 3 4 1 5 2
2
6 3 7 1 2 8 1 3 9 2 3
y X X X X X
X X X X X X X (4.34)
140
Where y in is the response at X1, X2 and X3. (Note Chapter 2, section 2.2.1
refers to for the basic information on the missing column matrix, along with the
calculation of the ANOVA terms.)
4.5.1 Mould Design
A mould, consisting of four cylindrical specimens was designed to
accommodate all combinations of factors, as per the above design table. The
cylindrical specimens branch out from the central sprue, as shown in Figure
4.5. The mould itself was a horizontally parted mould, with a 30mm long
runner, and mould walls shelled to a 5mm thickness. The mould design
provided two configurations, to account for the variation of the cavity angle
from the central sprue. Figure 4.5 below depicts the two configurations used.
The first design below, has the cavity angled at 0° and 45°. The second
configuration beneath, shows the cavity angled at 90° to the centre of rotation.
(a) (b)
Figure 4.5 CAD models of the centrifugal ZP131 plaster moulds used in the
experimental trials, showing (a) samples angles 90°to runner and (b) samples
orientated at 0 and 45° in relation to the runner bar.
4.5.2 Setup and Methodology
The CC experimental work was conducted on a custom built vertical CC
machine, located at Centracast Engineering, Auckland. The machine is
pneumatically operated and is able to control speed of rotation, pouring
141
speed, motor delay10 and ramp rate11. Castings with a diameter up to 720mm
and height up to 250mm can be accommodated, and the mould can be
preheated to the desired temperature. The operation12 of the machine is
controlled by a Graphical User Interface (GUI), which controls the machine
with the set parameters.
The original use of this CC machine was to utilise a steel die in which Al
would be cast. Steel dies machine were thus to attached to the machine
rotated by a connecting spindle. To incorporate a 3D printed mould into this
design, a two part hollow steel die with 300mm outside diameter was
fabricated from 1040 steel, inside which the 3D printed moulds could be
located. Cold setting easter hardened resin polymer coated sand was placed
around the 3D printed moulds to act as a rigid support for the mould and also
stop any metal spillage through the parting line. The steel die (metal
enclosure)13 shown below in Figure 4.6, was fabricated at AUT University
using a CNC machine.
Figure 4.6 CAD sectional model of the centrifugal die and mould setup
10
Time taken for the mould to rotate after the pouring has been completed
11 Acceleration of the mould up to the set rotation speed.
12 For detailed machine operation please refer to Appendix…
13 Refer to Appendix C for the die CAD drawings of the material information
142
Figure 4.7 Centrifugal casting setup showing steel die, pouring cup and holding
crucible
For the initial castings, Al A356 was poured at rotational speeds varying from
150RPM to 350RPM at temperatures from about 730°C. Initial testing‟s
trialled both the ZCast and ZP131 mould materials at different shell
thicknesses, baked at optimum time and temperatures. Before casting, the
printed moulds were preheated in a microwave oven at a temperature of
about 100°C. After two hours of preheating, the printed moulds were placed in
the steel enclosure and supported with the backing sand. Before pouring, the
steel die and pouring equipment were preheated with several gas torches to
remove any residual moisture. The metal was then transported in small
aluminium oxide crucibles from the furnace to the holding crucible on the
centrifugal casting machine, and then poured automatically.
4.5.3 Experimental Procedure
Following completion initial trials, the experimental trials were conducted in
accordance with Table 4.3. Each 3D printed mould was baked to the optimum
levels of permeability and compressive strength established from the material
testing covered in Chapter 2. The moulds were preheated at a baking
temperature of around 100°C for about an hour prior to casting. Before
casting, the preheated mould was placed in the locating centre hole in the
143
steel die shown in Figure 4.3 to position it directly under the down sprue of the
CC machine. Chemically bonded backing sand was then packed around the
3D printed moulds to act as a support for the mould and also prevent any
turning of the mould relative to the metal enclosure. This was achieved by
introducing holes into the bottom surface of the metal enclosure into which the
chemically bonded sand would form. After the supporting sand had bonded to
its full strength, the mould was placed on the CC machine and this was setup
for mould filling under CC operations. After casting the mould was allowed to
sit for twenty minutes before the casting was removed.
All castings were made in un-degassed CC601 alloy, which was melted in a
small electrical resistance furnace, at about 720°C. For casting the Al alloy
was melted in ceramic crucibles, which were charged with the CC601 ingot.
Before each melt, the crucible was cleaned of any oxide left from the previous
casting. Once the crucible was taken out of the furnace, the Al was poured
into a ceramic holding crucible in the CC machine, which was preheated to
around 200°C. The same pouring time was used in all the castings. The motor
delay was set to one second (the lowest value) for the castings performed at
150 and 300RPM and two seconds for the castings spun at 450RPM. The
desired amount of Al was poured into the holding crucible inside the CC
machine casting commenced. Every casting was spun for five minutes. After
this period, the top half of the metal die was lifted away, and the mould was
removed and allowed to air cool.
4.6 Centrifugal Casting Results
The experimental design and the responses measured with the CC trials
involving Al are presented in Table 4.4. Average values of each response are
plotted against rotational speed in Figure 4.8 for ease of comparison and a
detailed discussion of these results follows next.
144
Table 4.4 Overall results of mechanical testing of centrifugal test bars
Mould
Speed
(RPM)
Mould
Cavity
angle
(°)
Runner
cavity
ratio
Surface
roughness,
Ra
(μm)
UTS
(MPa)
Yield
strength,
0.2%
offset
(MPa)
Strain
%
150 45 1:1 13.27 92.60 77.00 0.95
450 45 1:1 18.15 128.60 87.25 1.49
150 180 1:1 13.05 99.60 82.90 1.13
450 180 1:1 19.07 109.55 93.45 0.93
150 90 0.8:1 15.24 106.90 85.70 1.33
450 90 0.8:1 14.31 103.20 91.50 0.86
150 90 1.2:1 11.98 104.10 88.30 0.75
450 90 1.2:1 11.62 107.30 82.90 1.11
300 45 0.8:1 13.17 102.70 79.00 1.31
300 180 0.8:1 13.54 106.00 82.35 1.09
300 45 1.2:1 14.78 116.70 85.95 2.10
300 180 1.2:1 14.10 112.50 87.80 1.15
300 90 1:1 12.22 120.60 89.10 1.47
300 90 1:1 12.22 109.80 80.90 1.44
300 90 1:1 12.53 144.55 95.20 2.35
(a)
8.009.00
10.0011.0012.0013.0014.0015.0016.0017.0018.00
150 RPM 300RPM 450RPM
Surf
ace
rou
ghn
ess
(μm
)
145
(b)
(c)
Figure 4.8 Graphical form of overall results from testing at various rotational speeds,
specifically (a) surface roughness (b) ASTM grain size of primary dendrites and (c)
tensile properties14
.
4.6.1 Surface Roughness
The ability of these moulds to withstand the velocities and pressures created
upon spinning was not previously known and it was suspected, there may be
some increase in surface roughness values due to mould erosion resulting
from turbulence, or perhaps even mould cracking due to the larger forces
acting. However, the moulds withstood the rotational forces in all cases and
only some flash was observed between the upper and lower halves of the
moulds, which is a common phenomenon in CC, due to the increased
14
Compiled averages from each speed range
2.25
2.30
2.35
2.40
2.45
2.50
2.55
2.60
150 RPM 300RPM 450RPM
AST
M g
rain
siz
e
0.80
1.00
1.20
1.40
1.60
0.00
20.00
40.00
60.00
80.00
100.00
120.00
140.00
150 RPM 300RPM 450RPM
% s
trai
n
Stre
ss (
MP
a)
Yield Strength UTS % elongation
146
pressures developed in the metal. While the use of the rigid backing sand
provided resistance to mould cracking, the internal surfaces of the cavity
exhibited no visual signs of mould surface erosion. Also, there was no
evidence of mould surface erosion present on the cast specimen surface,
indicating that mould erosion was not a significant problem with 3D printed
moulds when for CC.
In general, the overall surface roughness results were acceptable, with the
averaged values ranging from 13-15.5μm over the tested speed ranges.
These values appear to be slightly higher than those achieved in the static
casting trials (5.84μm), but it is noted that the ZP131 plaster moulds used in
the centrifugal casting trials were uncoated. This was done as the cast
surfaces actually gave a more consistent finish, due to the absence of the
markings induced by painting the mould coating onto the mould surface. This
resulted in well defined shapes and sections with a consistent surface finish.
Upon initial visual inspection and touch comparison, the cast surface
roughness was very comparable to those found from Al sand castings
conducted at the company using chemically bonded silica sand.
From the overall mechanical results in Figure 4.8, the surface roughness was
relatively stable at rotational speeds of 150 and 300 RPM respectively. The
results obtained at 450 RPM however, were elevated by around 2.5μm over
those the two lower speed settings. Without testing at higher speeds, it cannot
be concluded that this was due to increased mould erosion resulting from the
higher rotational speeds. But given that all the other factors were relatively
constant, such as mould production method, baking time, melt temperature
and method of pouring, it seems likely that this increase was influenced by the
higher rotational speed. The mechanism for this increase in cast surface
roughness could be due to either increased mould erosion at higher speeds or
increased metal penetration due to higher pressures present within the liquid
metal. Figure 4.9 below depicts the effect of metal penetration at (a) low
pressure and (b) high pressure.
147
(a) Low pressure
(b) High pressure
Figure 4.9 Possible influence of pressure on surface roughness at (a) low & (b) high
pressures
The increased pressure forces the metal into the mould surface, overcoming
the surface tension of the liquid metal. This may have resulted in the metal
forming a profile closer to the mould surface profile, leading to an increase in
the average surface roughness.
148
(a) 150 RPM
(b) 450 RPM
Figure 4.10 Typical cast surfaces at (a) 150 RPM and (b) 450 rpm
From Figure 4.10, it can be seen that the surface of the specimen cast at 450
RPM was considerably rougher than that of the casting spun at 150RPM. The
presence of greater pressure was also confirmed by the occurrence of a large
flash around the outer surface of the cylinder. However, this increased
pressure did not seem to affect the surface roughness values of the castings
conducted at 150 and 300RPM. Perhaps this occurred at only the highest
rotational speeds (450RPM) as the pressure generated overcame the
threshold of the surface tension force, pushing the metal into the small peaks
and valleys on the mould surface.
149
The other mechanism which may cause rougher surfaces could be the
increased velocity of the melt flowing over the mould surface. As the mould is
rotated at higher speeds, the shear stress present at the mould wall (due to
the flowing metal) will increase as a function of the square of the melt velocity.
This shearing action may dislodge gypsum plaster particles at the surface
leading to some loose mould material „washing through‟ with the melt and
leading to rougher cast surfaces. The equation below shows the relationship
between shear stress present at the wall and flow velocity of a given fluid [96]:
2
0
f V
8 (4.34)
Where:
0= Shear stress at the mould wall (N/m2)
f = friction factor
= Fluid density (kg‟s/m3)
V = Fluid velocity (m/s)
Equation (4.34) shows that the flow velocity is the dominating variable with
respect to the shear stress experienced at the wall. Again however, there is
no significant difference in the surface roughness between 150RPM and
300RPM, though there is considerable increase in the fluid velocity. There
may be however a critical value of the shear stress, of the gypsum particles,
which is overcome at higher rotational speeds above 300RPM. Figure 4.8
clearly shows that cast surface roughness was highest at 450RPM with
respect to lower speeds of 150 and 300RPM.
The empirical model built based on the surface roughness values measured
in the experiment is given below as equation (4.35). Figure 4.11 presents
plots showing variation in surface roughness with varying factors and it can be
seen that the effect of increased cavity angle results in reduction of the overall
surface roughness over the tested speed range (150-450RPM). As the cavity
angle increases the surface roughness reduces because of reduced melt
velocity. As the fluid flows around a bend, this acts as an obstruction to flow,
150
reducing the fluid velocity. The sharper the angle the greater this loss
becomes, resulting in a reduction in the surface roughness. The effect of the
runner to cavity ratio (r/c) seems to be constant for ratios ranging from 0.8-
1.2. Although changing the runner diameter should result in velocity change
(as the flow rate is the product of area and velocity) this does not seem to
have any effect on the surface roughness of the cast parts. However, the total
variation in the r/c ratio was small and might not have had a significant effect
on velocity of filling of the liquid metal, both generally and in particular on the
outer layers, where surface roughness is affected.
1 2 3 1 2
2 2 2
3 1 2 3
1 2 1 3 2 3
( , , ) 15.7924 0.0436 0.142
1.3237 0.0000656 0.000511 0.498
0.0000739 0.000293 0.000344
Ra X X X X X
X X X X
X X X X X X (4.35)
Where:
X1, X2, and X3 are rotational speed, cavity angle and r/c, respectively
The angle and the square of the cavity angle (X2 and X22) were the most
significant parameters at 73.9 and 85.5% confidence respectively (Table 4.5).
For the squared cavity angle this confidence level was close to the standard
level and is a near statistically significant result. The square of the speed term
was meaningful with a 73.8% confidence level. The other model terms,
specifically the interaction terms and linear and square runner diameter had
little influence on the overall model output. These effects however may be
significant at other levels outside this experiment and it cannot be assumed
that these effects vary linearly.
151
(a) 0.8:1 (b) 1:1
(c) 1.1:1 (d) 1.2:1
Figure 4.11 Graphical plots of surface roughness model results at runner-cavity (r/c)
ratios of (a) 0.8:1 (b) 1:1 (c) 1.1:1 and (d) 1.2:1.
10
12
14
16
18
20
0 200 400 600Surf
ace
ro
ugh
nes
s (μ
m)
Rotational speed (RPM)
0 degrees 45 degrees
90 degrees
10
12
14
16
18
20
0 200 400 600Surf
ace
ro
ugh
nes
s (μ
m)
Rotational speed (RPM)
0 degrees 45 degrees
90 degrees
10
12
14
16
18
20
0 200 400 600
Surf
ace
ro
ugh
ne
ss (μ
m)
Rotational speed (RPM)
0 degrees 45 degrees
90 degrees
10
12
14
16
18
20
0 200 400 600Surf
ace
ro
ugh
nes
s (μ
m)
Rotational speed (RPM)
0 degrees 45 degrees
90 degrees
152
Table 4.5 Contribution and significance of individual terms for the surface response
model
Term Co-eff. SE Co-
eff.
T P
Constant 15.7924 33.2018 0.476 0.654
Speed -0.0436 0.0502 -0.867 0.426
Angle -0.142 0.112 -1.268 0.261
Runner diameter 1.3237 3.7651 0.352 0.739
Speed 2 0.0000656 0.0001 1.263 0.262
Angle2 0.000511 0.0003 1.727 0.145
Runner diameter 2 -0.0498 0.1142 -0.436 0.681
Speed x Angle 0.0000739 0.0001 0.685 0.524
Speed x Runner diameter 0.000292 0.0023 0.125 0.905
Angle x Runner diameter 0.000311 0.0051 0.068 0.948
Table 4.6 ANOVA table showing linear, square and interaction model effects
Source DF Seq SS Adj SS Adj MS F P
Regression 9 40.5064 40.5064 4.50071 0.89 0.586
Linear 3 14.3204 13.3357 4.44523 0.88 0.511
Square 3 23.7147 23.7147 7.9049 1.57 0.308
Interaction 3 2.4713 2.4713 0.82377 0.16 0.917
Residual Error 5 25.2535 25.2535 5.05069
Lack-of-Fit 3 25.19 25.19 8.39667 264.68 0.004
Pure Error 2 0.0634 0.0634 0.03172
Total 14 65.7598
R2 = 61.6%
The ANOVA presented above in Table 4.6 revealed no significant factors at
90% confidence or above. The ANOVA table revealed that the regression
analysis conducted was not significant and only 61.6% (R2 term) of the
variation in the surface roughness was accounted for by the variation in the
experimental factors, in this case rotational speed, cavity angle and runner
153
diameter. The square portion of the regression accounted for the largest
variation, with smaller effects seen from the linear terms and especially the
interaction term.
The low R2 value was accompanied by a significant lack of fit test. This
showed that at a confidence level of 99.6 %, polynomial expression was not
an adequate model for the surface roughness response. The inability to fit the
data highlights that the varying factors were not affecting the response in a
way that a second order model could predict.
4.6.2 Yield Strength
The yield strength model developed is given in equation (4.36). The averaged
yield strength results (0.2% strain), shown below in Figure 4.12, showed a
slight increase at higher rotational speed. In particular, values increased
2.28MPa from 150-300RPM and 3MPa from 300-450RPM. This was a
significant result as the experimental values increased with rising rotational
speed. The yield strengths were however lower than those typically found in
traditional sand castings (90 MPa) [55].
1 2 3
2 2 2
1 2 3
1 2 1 3 2 3
8.31601 0.10124 0.201272 6.00576
0.00000166667 0.000989712 0.130615
0.0000865497 0.0058383333 0.00171784
yield X X X
X X X
X X X X X X (4.36)
154
Figure 4.12 Yield strength variation with rotational speed
The reasons for these lower values were attributed to larger grain sizes and
hydrogen porosity of the cast Al. In particular, the cooling rate of Plaster
(ZP131) moulds with their characteristic poor thermal conductivity, result in
grain growth and coarse grain structures at lower speeds. Further, the
dendrite cell size and spacing are linked with the yield strength through the
Hall-Pitch equation. Increases in grain size and DAS, thus decrease yield
strength of the cast parts.
For grain sizes to be decreased, and yield strength enhanced, increased
pressures during filling are necessary as this forces the metal against the
mould walls, achieving greater surface area of contact. This increases the rate
of heat transfer, resulting in a decrease in the solidification time and giving
slightly smaller dendrites. Also the eutectic structure is refined by the increase
in the cooling rate and this results in a change in morphology of these
particles, resulting in a more fibrous eutectic [88]. Figure 4.13 below shows
the differences in the dendritic and eutectic phases in static and centrifugally
cast A356 alloy. The static casting photomicrograph shows coarser grain
growth, whilst the centrifugally cast specimen shows a more refined structure.
80.00
81.00
82.00
83.00
84.00
85.00
86.00
87.00
88.00
89.00
90.00
150 RPM 300RPM 450RPM
Stre
ss (
MP
a)
Yield Strength Linear (Yield Strength)
155
(a) Static ZP131 casting (b) Centrifugal ZP131 casting
Figure 4.13 Comparison between (a) Static casting and (b) Centrifugal castings
produced in ZP131 plaster moulds (both X50)
The other possible reason for increase in yield strength at higher speeds was
increased pressure filling of the mould cavity. As highlighted in section 3.6.4,
Fe intermetallics influence the solidifying structure of Al castings, which has a
detrimental effect on cast strength. In static casting, these intermetallics block
the liquid eutectic mixture being fed through the dendrite network, resulting in
poor formation of dendrite cells and random placement of these particles [98].
Feeding has also been found to be drastically reduced when Fe intermetallics
are longer than a critical value [98]. In addition, the β phase Fe intermetallics
have severe effects on feeding ability and can often lead to porosity formation.
These β type partials were observed in the CC in the current work.
Perhaps under increased pressure at higher rotational speeds, the eutectic
liquid was able to be forced through the dendrite network more effectively,
resulting in fewer defects and a more sound grain structure, with little
shrinkage and fewer discontinuities. However, the presence of large porosities
due to hydrogen in the melt leads to below average yield and UTS compared
to traditional sand casting.
300μm 300μm
156
(a) 0.8 (b) 1
(c) 1.1 (d) 1.2
Figure 4.14 The Graphical plots of yield strength model at r/c ratios of (a) 0.8 (b) 1 (c)
1.1 and (d) 1.2
From the plots of Figure 4.14 clear trends were observed with respect to
rotational speed, cavity angle and runner-cavity diameter ratio. Firstly, the
rotational speed had a positive linear effect on the yield strength over the
entire speed range at each cavity angle, for r/c ratios ranging from 0.8-1.1.
For r/c ratio of 1.2 however, yield strength was seen to plateau and then
decrease with respect to all three cavity angels. Secondly, increasing cavity
angle resulted in increased yield strength, with 45° and 90° angles giving the
lowest and highest values of yield strength respectively, with the 0° angle
between these two results.
70
75
80
85
90
95
100 150 200 250 300 350 400 450 500
Yie
ld s
tren
gth
(M
Pa)
Rotational speed (RPM)0 degrees 45 degrees
90 degrees
70
75
80
85
90
95
100 150 200 250 300 350 400 450 500
Yie
ld s
tren
gth
(M
Pa)
Rotational speed (RPM)0 degrees 45 degrees
90 degrees
70
75
80
85
90
95
100 150 200 250 300 350 400 450 500
Yie
ld s
tren
gth
(M
Pa)
Rotational speed (RPM)
0 degrees 45 degrees
90 degrees
70
75
80
85
90
95
100150200250300350400450500
Yie
ld s
tren
gth
(M
Pa)
Rotational speed (RPM)
0 degrees 45 degrees
90 degrees
157
Table 4.7 Breakdown of model parameters in the yield strength model
Term Co-eff. SE Co-eff. T P
Constant 8.31601 93.4441 0.089 0.933
Speed 0.101237 0.1414 0.716 0.506
Angle 0.201272 0.3151 0.639 0.551
Runner diameter 6.00576 10.5966 0.567 0.595
Speed 2 1.66667E-06 0.0001 0.011 0.991
Angle2 -0.00098712 0.0008 -1.188 0.288
Runner diameter 2 -0.130615 0.3215 -0.406 0.701
Speed x Angle 0.0000865497 0.0003 0.285 0.787
Speed x Runner diameter -0.00583333 0.0066 -0.885 0.417
Angle x Runner diameter 0.00171784 0.0143 0.121 0.909
With respect to the individual model parameters, no statistical significance, was
observed
Table 4.7. The square of the cavity angle however was contributing more than
any of other terms, at a confidence level of 71.2%. There was little effect from
the other factors and small interaction effects highlighted the lack of any
importance of factors undergoing simultaneous change.
Table 4.8 ANOVA table of the yield strength model response
Source DF Seq SS Adj SS Adj MS F P
Regression 9 178.67 178.67 19.85 0.5 0.830
Linear 3 82.4 39.3 13.1 0.33 0.806
Square 3 61.09 61.09 20.36 0.51 0.693
Interaction 3 35.18 35.18 11.73 0.29 0.829
Residual Error 5 200.03 200.03 40.01
Lack-of-Fit 3 97.05 97.05 32.35 0.63 0.662
Pure Error 2 102.98 102.98 51.49
Total 14 378.71
R2 = 47.18%
158
The ANOVA presented in Table 4.8 showed no significant relationships
between yield strength and the linear, square and interaction effects of the
model. The square terms however accounted for more variance than linear
and interaction portions. The R2 value was 47.18%, meaning, that over half of
the variation in the results was not accounted for by the model effects.
Despite this there was no significant lack of fit and the model was adequate in
fitting the yield strength data. The regression itself was not significant as the
overall residual sum of squares (SS) (values that cannot be accounted for)
was relatively large (200.03) compared to the total SS (378.71). These low
confidence values may indicate a large standard deviation. This leads to the
conclusion that other significant variables may be involved to explain the
variance, such as porosity level.
4.6.3 Ultimate Tensile Strength (UTS)
From the averaged results shown in Figure 4.8c, there was an increase in the
UTS over the speed range. The tested UTS values ranged from 90.5-
145.8MPa for non-degassed CC601 alloy. The averaged values showed a
16MPa increase from 150RPM to 300RPM and a 12MPa increase from 150 to
450RPM. The middle speed setting gave the highest average UTS values
with an average of 116MPa. Although there was a slight decrease in the
average strength from 300-450RPM, the overall linear trend was positive with
respect to rotational speed. The results from the analytical model in section
4.3.2 indicate that these higher rotational speeds result in relatively higher
pressures, up to 300mmAl. Simple T-tests conducted between the different
speeds are shown below in Table 4.9, clearly indicating that the increases in
UTS at each speed were statistically significant.
Table 4.9 Results of T-Tests conducted between each speed range.
T-Test Confidence level from T-Test
150 and 300 RPM 97%
150 and 450 RPM 86%
300 and 450 RPM 52%
159
High confidence intervals were established between 150 and 300 and 150
and 450RPM in relation to the statistical difference of their averages (Table
4.9). A statistically significant result is seen when comparing speed at
300RPM with a near statistically significant result at 450RPM. (Note: the
importance of these results should not be confused with the ANOVA model
which takes into account linear, square and interaction effects contributing to
the overall model variance).
When compared to values achieved from the static castings, much more
variability in the UTS of the CC cast specimens was observed. Both the
lowest and highest UTS values were achieved in the CC process, indicating
considerable process influence on the final cast strength.
The UTS values relied greatly on the size and distribution of the porosity, due
to the dissolved hydrogen in the melt. Due to restrictions in the casting setup
used, degassing of the molten metal was unable to be incorporated, leading
to high levels of hydrogen in the molten metal. When examining the tensile
specimens, large pores were observed but these sizes were seen to reduce
when rotational speed was increased to 300 and 450RPM respectively.
The size and distribution of these pores were dominant in reducing the
strength of these cast samples. Lesser effects such as particle inclusion and
oxide entrapment do provide an effect by reducing the strength developed in
the microstructure (hydrogen porosity was covered in section 3.6.4). It was
suspected that correct melt treatment (degassing) before pouring would yield
significant improvements in cast strength. To confirm this, a final CC was
conducted at 235RPM at AUT University using a modified setup, with samples
angled at 0 and 45°respectively. The melt was lance degassed with Nitrogen
for three minutes before pouring and the tensile properties are listed below
(refer to 4.8.2 for methodology).
160
Table 4.10 Tensile properties of centrifugally cast degassed specimens
Sample UTS (MPa) Yield Strength
(0.2% offset)
% elongation
0° 154 118.5 2.03
45° 151.4 121.1 1.95
This result shown above in Table 4.10 was very significant as large
improvements in yield and UTS values resulted. It is noted that lance
degassing is not very efficient at removing the dissolved hydrogen when
compared with other degassing methods, such as rotary degassing [99].
Despite this, the highest tensile strength values were achieved with the
degassed samples. Further, the rotational speed was not as high as
previously used (Centracast trials).
For the Centracast trials, (un-degassed), the model developed for the UTS is
given in equation (4.37). The plots of the UTS in Figure 4.15, showed that all
three variables had an influence on UTS values. The overall trend was an
initial increase of strength from 150RPM to about 250-350RPM, followed by a
slight decrease up to the maximum speed, 450RPM. Again, the effect of
increasing cavity angle resulted in the raising of the strength values from
angle change 45° to 90°. The 0° angle however resulted in generally lower
strengths obtained at higher speeds ranging from 350-450RPM.
1 2 3
2 2 2
1 2 3
1 2 1 3 2 3
184.070 0.310348 0.519288 27.9396
0.00047768 0.00150283 0.865275
0.000428265 0.00359375 0.00471491
UTS X X X
X X X
X X X X X X (4.37)
161
(a) 0.8:1 (b) 1:1
(c) 1.1:1 (d) 1.2:1
Figure 4.15 The Graphical plots of Ultimate Tensile strength model at r/c ratios of (a)
0.8:1 (b) 1:1 (c) 1.1:1 and (d) 1.2:1.
A decrease in UTS was observed at high rotational speeds when compared to
the average value obtained at 300RPM. Although this suggests improved
properties at 300RPM, this result was somewhat skewed by a large strength
value in trial 15 (Table 4.4) at 300RPM, which gave UTS of 144.55 MPa. If
this result was excluded, the average UTS at 300RPM and 450RPM become
similar. To illustrate this, Figure 4.16 shows the average UTS values with trial
15 excluded.
85.00
90.00
95.00
100.00
105.00
110.00
115.00
120.00
100 200 300 400 500
Ult
imat
e te
nsi
le s
tren
gth
(M
Pa)
Rotational speed (RPM)
0 degrees 45 degrees
90 degrees
85.00
95.00
105.00
115.00
125.00
135.00
100 200 300 400 500
Ult
imat
e te
nsi
le s
tren
gth
(M
Pa)
Rotational speed (RPM)
0 degrees 45 degrees
90 degrees
85.00
95.00
105.00
115.00
125.00
135.00
100 200 300 400 500
Ult
imat
e te
nsi
le s
tren
gth
(M
Pa)
Rotational speed (RPM)
0 degrees 45 degrees
90 degrees
85.00
95.00
105.00
115.00
125.00
135.00
100 200 300 400 500
Ult
imat
e te
nsi
le s
tren
gth
(M
Pa)
Rotational speed (RPM)
45 degrees 90 degress
0 degrees
162
Figure 4.16 UTS values of specimens ignoring results of trial 15
Table 4.11 Contribution of individual model parameters in the UTS model
Term Co-eff. SE Co-eff. T P
Constant -184.07 215.499 -0.854 0.432
Speed 0.310348 0.326 0.952 0.385
Angle 0.519288 0.727 0.715 0.507
Runner diameter 27.9396 24.438 1.143 0.305
Speed 2 -0.00047768 0 -1.416 0.216
Angle2 -0.00150823 0.002 -0.785 0.468
Runner diameter 2 -0.865275 0.741 -1.167 0.296
Speed x Angle -0.000428265 0.001 -0.611 0.568
Speed x Runner diameter 0.00359375 0.015 0.237 0.822
Angle x Runner diameter -0.00471491 0.033 -0.143 0.892
With respect to Table 4.11, there were no statistically significant co-efficients
found for equation (4.38) but some medium level confidence values were
obtained, with the square of the rotational speed giving a 78.4% confidence
level. Also, linear and square terms of the r/c ratio were confident at 69.5%
and 70.4%, respectively. Lower confidence levels were observed for all
interaction effects, together with linear speed and cavity angle parameters.
100.80
113.20 112.60
90.00
95.00
100.00
105.00
110.00
115.00
120.00
150 RPM 300RPM 450RPM
Stre
ss (
MP
a)
UTS
163
Table 4.12 ANOVA of the UTS model response
Source DF Seq SS Adj SS Adj MS F P
Regression 9 1203.29 1203.29 133.7 0.63 0.744
Linear 3 362.71 453.92 151.31 0.71 0.586
Square 3 744.91 744.91 248.3 1.17 0.409
Interaction 3 95.67 95.67 31.89 0.15 0.925
Residual Error 5 1063.87 1063.87 212.77
Lack-of-Fit 3 431.27 431.27 143.76 0.45 0.742
Pure Error 2 632.6 632.6 316.3
Total 14 2267.16
R2 = 53.07%
The variance analysis shown in Table 4.12 also showed no statistically
significant linear square or interaction effects (P value <0.1). The square
portion of the model accounted for slightly more variation than the linear
model, with a very insignificant interaction effect. The sum of squares (SS) of
the linear portion for example was 362.71 and with the addition of the square
effects this increased to 744.91.
The overall regression model was only able to account for 53.07% of the total
variation in the experiment, leaving a large experimental residual term.
Nevertheless, no statistically significant lack of fit of the model was found,
showing that the chosen model was able to fit the data points. This highlights
that the large residual term must be due to the absence of some other
influencing factor. It has already been suggested that this may be hydrogen
porosity.
When conducting simple T-Test on the difference of the averages of un-
degassed (Centracast) specimens at 300RPM and the degassed specimens
(AUT University) at 235RPM a significance of 98.2% was observed,
confirming the effect of degassing was very considerable on the overall UTS
results.
164
4.6.4 Percent Elongation
The overall elongation values shown below in Figure 4.8c, were low and this
is due to characteristic brittle fracture. The elongation values ranged from
0.43-2.44%, with typically larger values linked to specimens with high UTS
values. The porosity present affected the ductility of the material as these
voids created stress concentrations, which led to premature fracture. Also, the
slower cooling rate might increased grain growth, leading to slightly coarser
dendrite and eutectic structures which provide less resistance to grain
boundary sliding.
Over the tested speed ranges, increases in ductility were observed with a
notable increase at the 300RPM speed. Interestingly, there was only a minor
difference of 0.06% strain between the samples tested at 150RPM and
450RPM, despite there being a 12MPa difference in UTS. At 300RPM the
elongation was increased to an overall average of 1.56%, considerably higher
than at the other two speed levels. Disregarding trial 15, which gave high
tensile test results, the average elongation was still 1.43%. When a T-Test
was conducted on the average elongation values over the tested speeds
there was no significance between 150RPM and 450RPM. However,
significant confidence levels of 94.8 and 91.8% were found at 300RPM as
against 150 and 450RPM respectively.
1 2 3
2 2 2
1 2 3
1 2 1 3 2 3
8.31610 0.101237 0.201272 6.00576
0.0000016 0.00098712 0.130615
0.0000865 0.005833 0.00171784
X X X
X X X
X X X X X X (4.38)
The elongation model was similar to that of the UTS model in terms of the
overall relationships and trends. The elongation values tended to increase
over the speed of 150-350RPM but decreased at higher rotational speeds
from about 350-450RPM. For increasing r/c ratios, all elongation values at
both the 45 and 90° angles increased initially from 150-350RPM, reached a
maximum at about 350RPM and then dropped but to values above those at
the 150RPM speed. The 0° angle at these higher r/c ratios tended to plateau
165
at lower speeds of about 250RPM and then proceeded to fall over speeds of
250-450RPM. At r/c ratios of 1 and 1.1, elongation values at the 0° angle
dropped below those found at 150RPM.
(a) 0.8:1 (b) 1:1
(c) 1.1:1 (d) 1.2:1
Figure 4.17 Individual plots of rotational speed and cavity angle with respect to the
percent elongation model at r/c ratios of (a) 0.8:1 (b) 1:1 (c) 1.1:1 and (d)1.2:1.
0.50
1.00
1.50
2.00
2.50
100 200 300 400 500
Elo
nga
tio
n (
%)
Rotational speed (RPM)
0 degrees 45 degrees
90 degrees
0.50
1.00
1.50
2.00
100 200 300 400 500
Elo
nga
tio
n (
%)
Rotational speed (RPM)
0 degrees 45 degrees
90 degrees
0.50
1.00
1.50
2.00
100 200 300 400 500
Elo
nga
tio
n (
%)
Rotational speed (RPM)
0 degrees 45 degrees
90 degrees
0.50
1.00
1.50
2.00
100 200 300 400 500
Elo
nga
tio
n (
%)
Rotational speed (RPM)
0 degrees 45 degrees
90 degrees
166
Table 4.13 Significance of the individual model parameters in the percent elongation
model
Term Co-eff. SE Co-eff. T P
Constant -5.58457 6.29297 -0.887 0.415
Speed 0.00855246 0.00952 0.898 0.41
Angle 0.0140219 0.02122 0.661 0.538
Runner diameter 0.656933 0.71363 0.921 0.4
Speed 2 -0.0000227889 0.00001 -2.313 0.069
Angle2 0.0000121399 0.00006 -0.216 0.837
Runner diameter 2 -0.0220459 0.02165 -1.018 0.355
Speed x Angle -0.0000154893 0.00002 -0.757 0.483
Speed x Runner diameter 0.000430208 0.00044 0.97 0.377
Angle x Runner diameter 0.000593933 0.00096 -0.619 0.563
The parameters in the model percent elongation expression (equation 4.38)
shown in Table 4.13, showed a statistically significant effect from the square
value speed term. This gave a 93.1% confidence level, which is statistically
significant. The other model parameters had relatively low confidence levels,
mostly in the range of 50-65% confidence.
Table 4.14 ANOVA of the various effects present in the percent elongation model.
Source DF Seq SS Adj SS Adj MS F P
Regression 9 1.7962 1.7962 0.19958 1.1 0.484
Linear 3 0.3526 0.2964 0.09879 0.54 0.673
Square 3 1.0997 1.0997 0.36657 2.02 0.23
Interaction 3 0.3439 0.3439 0.11464 0.63 0.626
Residual Error 5 0.9072 0.9072 0.18144
Lack-of-Fit 3 0.371 0.371 0.12368 0.46 0.738
Pure Error 2 0.5362 0.5362 0.26809
Total 14 2.7034
R2 = 66.44%
167
The ANOVA presented in Table 4.14 contained no significant linear, square or
interaction terms, though the addition of the square effects term did yield the
highest confidence level of 77%. The elongation model was the most accurate
model with a R2 value of 66.44%. This meant that two thirds of the total
experimental variation was accounted for by the regression analysis. There
was no significant lack of fit of the model. Nevertheless, the chosen factors
did not account for enough variation for the regression to be significant. The
sensitivity of the elongation values to the random distribution of the pores
present in the casting was a likely contributor to the unexplained variance
present in the results.
4.7 Macro and Micro Structural Examination
4.7.1 Macrostructures
The macrostructures of CC showed severe gas porosity throughout the cross-
sections of the samples. These pores were often distributed over the cross-
sectional area, indicating that these were not shrinkage pores but more likely
hydrogen gas entrapped by solidifying dendrites. The pore morphology in
Figure 4.18 shows the effect of higher rotational speeds on pore size.
The beneficial effect of higher rotational speeds also reduced the size and
distribution of these voids as increased pressures reduced pore size. The
mechanism for this centres on the initial and final volumes of a bubble due to
pressure change. By using Boyle‟s law the change in size of the bubble can
be determined, i.e.:
Final bubble volume
12 1
2
Pv v
P (4.49)
168
Where:
v = Volume (kg/m3)
P = Pressure (Pa)
(*Subscripts 1 and 2 denote initial and final conditions)
(a) 150RPM (b) 300RPM
(c) 450RPM
Figure 4.18 Macrostructures of cast specimens at (a) 150 (b) 300 and (c) 450 RPM
From equation (4.49), increasing fluid pressure will result in the reduction of
the final volume of the gas pore. With respect to Figure 4.18, larger and fewer
pores were seen at 150RPM, whereas at 300 and 450RPM much smaller and
evenly distributed pores were observed. The analytical model presented in
section 4.3.1 shows that these higher rotational speeds correlate to increased
pressures ranging from 150-300mmAl, as opposed to pressures of 135-
150mmAl at 150RPM.
169
The shape of these pores was also significant and generally these pores were
round, which is a characteristic of pores resulting from hydrogen porosity
[100]. The round pores seen in the macrostructures, particularly at speeds of
300RPM and 450RPM form by precipitating in the liquid metal at the start of
solidification [100]. This results in unrestricted bubble growth and is
characterised by Al containing high levels of hydrogen [100]. It seems
reasonable to conclude that the size and distribution of these pores formed
during CC were reduced by higher pressures present in the molten metal and
that this correlates to improved tensile properties at higher rotational speeds.
4.7.2 Microstructures
The as-cast microstructures presented in Figure 4.19a-e show typical dendrite
and eutectic phases expected with Al casting alloys containing 7% Silicon
(Si). Other inclusions, such as oxide skins and Fe intermetallics s also
observed in the samples. On average there was a slight decrease in the
dendrite grain size shown in Figure 4.8b when rotational speeds were
increased from 150 to 450RPM. The average grain number at 150, 300 and
450 RPM were found to be 2.38, 2.57 and 2.46 (ASTM grain number)
respectively. The lowest grain size was achieved at 300RPM, which also
correlates to the highest average UTS.
The overall microstructures did not show any unusual or modified phases,
indicating that the only other possible influence on mechanical properties
outside of the model was due to porosity. Previous researchers [91, 92, 101]
have deduced links between porosity defects and tensile properties, with their
results highlighting significant decreases in strength and ductility as the level
of porosity increases.
However, in the current work there was also a link between solidification time
and the total porosity present in the sample. The cooling rate increases
slightly with rotational speed and this reduces pore size as the dendrites
restrict hydrogen mobility [76]. Research has shown that at a specific
hydrogen content, the increase in cooling rate decreases both the average
170
pore size and the total amount of porosity present in A356 alloys [89]. Other
work showed that volume percent of porosity does not define the mechanical
properties in itself, rather it is the maximum pore lengths and areas which
correlate to the tensile properties [102].
(a) (b)
(c) (d)
(d) (e)
Figure 4.19 Microstructures of centrifugal castings, spun at 150RPM at X100 and X400
magnifications respectively (all) for (a & b) Trial 1 – 45°, 1:1 runner ratio (c & d) Trial 7 –
90°, 1.2:1 runner ratio and (d & e) Trial 3 – 180°, 1:1 runner ratio.
175μm 45μm
171
(a) (b)
(c) (d)
(d) (e)
Figure 4.20 Microstructures of castings spun at 300RPM at conditions (a & b) Trial 11 –
45°, 1.2:1 runner ratio (c & d) Trial 15 – 90°, 1:1 runner ratio and (d & e) Trial 12 – 90°,
1:1 runner ratio.
175μm 45μm
172
(a) (b)
(c) (d)
(d) (e)
Figure 4.21 The Microstructures of castings spun at 450RPM at conditions (a & b) Trial
2 – 45°, 1:1 runner ratio (c & d) Trial 8 – 90°, 1.2:1 runner ratio and (d & e) Trial 4 – 180°,
1:1 runner ratio.
175μm 45μm
173
(a) (b) (c)
Figure 4.22 Microporosity present at the three different rotational speeds, namely (a)
150RPM (b) 300RPM and (c) 450RPM
Microporosity featured throughout the castings at different rotational speeds
and these are shown in Figure 4.22. Microporosity weakens the
microstructure, as noted above, and the presence of entrapped oxides also
influences the strengths of Al-Si castings alloys. Significant research by
Campbell [103-105] on the role of entrapped oxides during casting has led to
the discovery of double oxide films (bifilms). The inclusion of bifilms in the cast
structure is influential in the development of strength, as it creates locations
from which cracks can propagate. Bifilms are entrained into the casting by the
action of pouring, which then folds the surface film, creating two adjacent
faces which do not bond together [105]. The good thermal stability of oxides is
a problem in metal casting as they stay suspended in the molten metal for
extended periods of time. Further, oxides are capable of unfolding
themselves, resulting in a major crack in the cast microstructure. The
centrifugal action itself expected to move these particles towards the axis of
rotation, moving the heavier particles to the outer radius of the mould cavity.
However, a large presence of oxides was still observed, resulting in a
weakening of the test samples. Also, the longer cooling times associated with
casting into plaster moulds may have resulted in more unravelling or unfurling
of these oxide files [106]. Shown below in Figure 4.23 were several examples
of entrained oxide films observed from the photomicrographs of the current
investigation, which resulted in a thin planar cavity, facilitating crack initiation.
300μm
174
(a)
(b)
(c)
Figure 4.23 Photomicrographs of entrained oxide films in castings conducted at (a)
150RPM (b) 300RPM and (c) 450RPM (all X50 magnification)
The presence of these oxides was probably due to several reasons. The
melting procedure firstly resulted in the formation of a surface layer of oxide,
which in the current setup was difficult to remove. This led to oxides entering
the mould. However, further issues result from large uncontrolled fall height,
300μm
175
shown below in Figure 4.24, which resulted in some initial oxides being
entrained into the casting. This fall height however was unavoidable, due to
the setup of the Centracast CC machine. Lastly, there was no filter used to
remove any entrained films, resulting in these defects being carried through
and subsequently trapped in the solidifying structure. To reduce the presence
of these entrained films, future trials should be conducted with improved
melting practices to reduce oxide formation. The use of filters specifically
needs to be incorporated, not only to remove oxide films but other non-
metallic inclusions.
Figure 4.24 Die and machine setup, with fall height depicted
4.7.3 Fractography of Centrifugal Castings
When evaluating the fractured specimens, it was evident that the presence of
porosity was dominant on all of the fractured surfaces, indicating that failure
initiated from these voids. The fracture surfaces15 showed both regions of
brittle and ductile fracture, with typical dimple fracture characterising the
ductile failures and smooth featureless surfaces characterising the brittle
fractures. The mechanism for fracture during tensile testing often originates
from particle inclusions or from eutectic silicon particles [66]. The large
15
Refer to Appendix B.3
333mm
176
presence of porosity throughout the samples resulted in the initiation of
fracture at these defects. The created voids are localised areas of intense
stress, with stress levels reaching up to five times greater than that of the Al-
Si Matrix [107]. Further, the level of stress intensity is greater for voids located
at the outer surface of the test sample. This surface porosity causes the initial
fracture from which cracks would propagate and link with other internal pores
in the sample. These stress concentrations affect both the primary Al
dendrites but more importantly the eutectic Si, which is a much more brittle
phase [107].
(a) 150RPM (b) 300RPM
(c) 450RPM
Figure 4.25 SEM photographs of the fractured surface at (a) 150 (b) 300 and (c)
450RPM.
177
Evidence of porosity and other defects on the fractured surfaces was found
across all tested samples. Presence of the distribution of these pores was
evident from the results of the macrostructures, although SEM allowed more
vivid and detailed photographs of this phenomenon. The comparison between
samples cast at 150RPM, 300RPM and 450RPM is shown in Figure 4.25.
Evidence of oxide films was also present on the fractured surfaces and Figure
4.26 below shows a large film present on the surface of the tensile specimen.
(a)
(b)
Figure 4.26 SEM photographs showing (a) long oxide film and (b) Fe-intermetallic
particle
Entrapped
oxide
Fe
intermetallic
178
4.8 Further Centrifugal Casting
Most of the CC trials were conducted with Al alloys, as the current setup and
resources were not suitable to process Mg in this CC machine. Several
attempts were made at melting the Mg at Centracast, with gas fired torches
used to heat small steel crucibles containing AZ91 Mg alloy. However, this
was an inadequate heating method which seemed to cause excessive
reaction and oxidation. As a result, no castings were produced. A new casting
setup was then constructed at AUT University. An initial trial was conducted
with Mg to prove that these moulds were capable of being cast centrifugally
and to analyse mechanical properties of the CC Mg alloy.
4.8.1 An Improvised Centrifugal Casting Setup
The casting setup adopted an existing rotating table. An electrically controlled
Frema Mini potter‟s wheel shown in Figure 4.27, was chosen to rotate these
RP moulds. This machine was capable of rotating parts up to thirty kilograms
and speed can be varied from 0-235RPM. The direction of rotation was able
to be varied both clockwise and counter clock wise. The metal was poured
into the moulds whilst stationary and then a protective cover, such as a steel
enclosure, was placed over the mould and the machine was allowed to spin
for about two minutes.
Figure 4.27 Improvised potter’s wheel used for CC of Mg
179
4.8.2 Centrifugal Casting Trials with the New Setup
Aluminium trial
The same 3D printed moulds used in the previous CC trials were employed in
this new setup. A one off trial involving CC601 Al alloy was conducted at
235RPM to first ensure the proposed method was capable of producing a
casting. The 3D printed mould was held in place by a mixture of foundry green
sand and hardened gypsum plaster. Before casting the metal the melt was
degassed with commercially pure Nitrogen through the use of a lance. The
melt was degassed for about five minutes prior to pouring and the metal was
cast at 710°C. The CC trial was successful, thus allowing the equipment to be
trialled with the more dangerous Mg alloy.
Magnesium trials
Two Mg casting trials were conducted with AZ91 magnesium at rotational
speeds of 235RPM. The first trial was conducted at AUT University‟s metal
casting laboratory, in which the Mg was heated to 730°C and then poured into
the 3D printed mould shown in Figure 4.5b. The mould was kept stationary
and the alloy was poured into the mould. Once the mould cavity was filled, a
metal cover was placed over the rotating table, whilst a protective gas mixture
was applied through a small opening at the top of the steel enclosure. Power
was applied to the machine and the mould was spun for about two minutes.
Excessive oxidation of the Mg occurred, which culminated in a sudden
release of molten Mg upon removal of the pouring cup. Protective gas and
flux was placed on the reacting metal, but the Mg continued to oxidise
completely. The practical difficulties in containing Mg fires were apparent and
it was decided that future work, would need to be conducted offsite.
A local foundry, Progressive Castings obliged, and a final trial was conducted
onsite. The trial consisted of the same setup at AUT; however the Mg alloy
was treated somewhat differently. This involved melting the alloy under the
cover of a flux rather than a cover gas. Also, chemically bonded sand instead
180
of a greensand was used to hold the rotating 3D printed mould in position.
This sand was very similar to that used as the backing sand in the Al
centrifugal trials conducted at Centracast. The mould had a proprietary
coating applied to reduce mould metal reaction. Finally, before pouring, the
mould was flushed with a sulphur containing gas, and then the metal was
cast. The melt was cast at about 710°C and the same protective cover was
placed over the rotating mould. The casting was spun for two minutes. This
proved successful and initial tensile and microstructural evaluation was
conducted.
4.8.3 Results of the Modified Centrifugal Casting Trials
Aluminium castings
The tensile testing results on the cast parts produced in the modified CC
machine are shown in Table 4.15. Large increases in all tensile properties
were obtained with the highest UTS and yields strengths achieved, obtained,
from the castings produced from this modified setup. The average strengths
of these parts were 152.70 MPa and this was considerably higher than the
cast strengths achieved at higher rotational speeds, conducted at Centracast.
The difference in yield strength was also significant with an average of
116.95MPa, compared to 86.88MPa for the castings spun at 300RPM
(optimum setting) at Centracast. The difference in percent elongation to
earlier trials conducted at the company was not as pronounced, with an
average of 1.99% for the aluminium trials, compared to the optimum of 1.60%.
Table 4.15 Tensile testing results from centrifugal castings, produced with CC601
aluminium
Sample UTS
(MPa)
Yield Strength (0.2% off-
set) (MPa)
%
Elongation
Aluminium 0° 154.00 118.50 2.03
45° 151.40 115.40 1.95
181
The difference in fall height (about 250mm), and the use of lance degassing
are likely reasons why these castings were significantly stronger than those
trails conducted at Centracast. Although inefficient, the lance degassing would
have removed some portion of the dissolved hydrogen, resulting in a less
porous casting. However, upon turning down the cast samples to the required
test shape, porosity was still observed throughout the samples but high
strengths were obtained, indicating that efficient degassing should result in
even higher cast properties. The arguments presented for the low tensile
properties of the Centracast casting produced previously centred on the melt
quality, with dissolved hydrogen shown as the dominating adverse factor.
These arguments have been validated as lance degassing resulted in much
higher strengths.
Analysis of the microstructure in Figure 4.28 showed the typical modified
eutectic Si particles surrounded by the primary Al dendrites. Considerably
fewer pores were present, and there were significantly smaller than those
from un-degassed samples. The reduction in the number of pores and pore
size accounted for the higher tensile properties of degassed the Al alloy. The
sensitivity of the as-cast tensile strength to the quantity and morphology of
these pores cannot be underestimated. Even with modest lance type
degassing, the UTS were on average 10MPa higher than those samples cast
in chemically bonded foundry sand which also underwent rotary degassing
(best practice).
As referred to in Figure 4.24, the fall height of the metal in the Centracast CC
machine probably caused excessive entrainment of metal oxides into the
castings. However the lower fall height in this modified setup resulted in a
lower number of oxide films being present in the cast microstructure, resulting
in sounder castings.
182
(a) X50
(b) X100
(c) X400
Figure 4.28 Al as-cast microstructure of lance degassed centrifugal castings at (a) X50
(b) X100 and (c) X400
300μm
175μm
45μm
183
Mg castings
These Mg castings proved that the overall concept of CC of Mg by 3D printing
was a viable concept when cast with the help of industry professionals. No
significant oxidation, mould-metal reaction or mould deformation was
observed from this casting trial, leading the author to conclude that these 3D
printed moulds were indeed suitable for processing Mg alloys. The problems
faced in the initial trial at AUT University were primarily practical, relating to
melting and the excessive tendency for oxidation of Mg alloys. The casting
shown below is just after the Mg was poured
Figure 4.29 AZ91 CC shown in ZP131 printed mould (left) with close-up of cast surface
shown on the right
Surface testing results of the as-cast surfaces gave an average roughness of
19.16μm. This was slightly higher than the static casting produced under
similar conditions. The static casting produced with ZP131 mould material,
AZ91 alloy gave a surface roughness of 9.49μm, whereas testing on the
centrifugal cast surface ranged from 11.88-26.48μm. The application of the
mould coating may be partly due to these changes in surface roughness.
Before this CC trial was conducted, the 3D printed mould was coated by
foundry personal. Brush coating of moulds in the static casting trials showed
that much detail and care was taken to avoid surface deformation. These
practices were not evident here, due to the trial nature of the casting. Also,
excessive turbulence and vibration during CC may have contributed to mould
metal reaction, leading to surface porosity, increasing surface roughness. The
184
actual cast strength of the centrifugal Mg castings were low in relation to the
strengths achieved in the static Mg casting trials.
Table 4.16 Results of centrifugal magnesium castings
Sample UTS
(MPa)
Yield Strength (0.2% off-
set) (MPa)
%
Elongation
Magnesium 0° 113.00 100.80 1.22
45° 93.50 N/A 0.64
These tensile properties summarised in Table 4.16, ranged from 93-
113.50MPa, which was significantly lower than static casting trials, which
gave 156.42MPa average using the same mould materials. The ductility of
these castings was also low, with only 0.64-1.22% elongations obtained.
These figures indicate that foreign materials, such as the covering flux,
became entrained in the cast microstructure. On a macro level, when turning
down the specimens to the tensile test shape, the castings showed no
porosity or shrinkage.
The resulting lower tensile properties must be reflect by the as-cast
microstructure. As a result, a tentative microstructural analysis was conducted
on the cast parts and these findings are shown below. The cast microstructure
shown in Figure 4.30 revealed slightly refined grains when compared to those
in the static casting trials in ZP131 moulds. This refinement was a likely result
of an increased solidification rate brought on by CC effect. However, as noted
above, these advantages did not prevent the low tensile properties were
obtained.
Upon examination, the microstructures showed shrinkage and microporosity
was observed, possibly contributing to these poor tensile properties. The dark
areas in Figure 4.30a&b represent microporosity weakening the cast
structure. However, Figure 4.30c shows a slightly different eutectic structure,
where a more fully divorced structure obtained. As noted in Chapter 3, fully
divorced structures relate to higher solidification rates. In static casting results,
185
this did not translate to improved properties. These eutectic particles may be
flux material trapped in the solidifying structure, as these particles were not
observed elsewhere in the static casting Mg trials.
(a) X50
(b) X100
(c) X400
Figure 4.30 Photomicrographs of centrifugally cast AZ91 in ZP131 moulds at (a) X50 (b)
X100 and (c) X400
300μm
175μm
45μm
186
Figure 4.31 below showed severe example of shrinkage.
Figure 4.31 Suspected shrinkage porosity in AZ91 CC (X100)
4.9 Summary of Centrifugal Casting Trials
The CC trials using rotating moulds resulted in interesting results, which
varied over a large range of values. The samples, however, were badly
affected by the presence of hydrogen porosity, which became the dominant
failure mechanism in Al castings. Nevertheless, increasing rotational speed
was linked to increases in tensile properties such as grain number and tensile
strength. Also, it was found that increasing rotational speeds resulted in a
dramatic decrease in porosity pore size and distribution, which was closely
linked to improved material strength.
The additional trials substantially increased Al tensile properties, giving the
highest yield and UTS values in this entire work. Final trials, involving AZ91
Mg, alloy resulted in below average tensile properties. However, the suitability
of these RP moulds in processing Mg alloys was successful, in that hot
volatile Mg alloys were cast without mould deformation or excessive reaction.
175μm
187
It is conjectured that the low tensile properties of these Mg castings were a
function of the melt quality and not the RP mould used.
Further, there was no evidence of mould cracking or deformation under
increased filling pressures, with some castings producing superior as cast
tensile properties when compared to traditional sand castings. Overall, 3D
printed ZP131 plasters moulds used for CC were considered successful in the
context of the experimental constraints experienced and the problematical
variables not examined or not proven.
188
CHAPTER 5 CONCLUSIONS
It is submitted that the results of experimental investigations carried out in this
research were significant, and, specifically, that the suitability of RP moulds
for casting of both aluminium and magnesium was established. Critical
property data for different mould materials were developed for optimum
process conditions. Influences of several factors on the performance of RP
moulds while casting different alloys were outlined. The rapid produced
moulds were found to be suitable for both static and centrifugal casting and
for both aluminium and magnesium alloys. The following are the main
conclusions drawn from the results obtained:
Mould materials
Both and ZP131 ZCast501 were found to be suitable for casting light
metals namely, aluminium and magnesium.
The experimental factors selected had varying levels of significance on
performance of the RP mould materials.
For ZP131, the compressive strength model gave a statistically significant
regression analysis.
Significant effects were found in both permeability and compressive
strength models for the ZCast501 material.
Optimum baking times and temperatures for ZP131 were 150°C and 8
hours for the best compressive strength and 150°C and 3 hours for the
optimum permeability.
Optimum baking times and temperatures for ZCast501 were 174°C and
5.27 hours for the best compressive strength and 209°C and 7.70 hours
for the optimum permeability.
Post baking compressive strengths in both materials were comparable to
those using traditional sand casting materials at the optimum baking
conditions.
Permeability was low for both RP materials employed relative to values for
traditional foundry sand.
189
Static casting trials
Both aluminium and magnesium were cast successfully using RP moulds
Mechanical properties such as surface roughness, tensile strength,
percent elongation and hardness were optimised by selected factors,
namely mould material, mould coating, alloy type and pouring
temperature.
Based on Taguchi signal processing and variance analysis, the
significance and best combinations of the factors and their levels were
found for each response.
Specifically, ZP131 moulds were best suited for processing Al and Mg
mechanical properties, using a combination of magnesium oxide based
mould coating and a pouring temperature of 690°C.
The mechanical properties of static castings produced using RP moulds
were close to values reported in the traditional sand casting literature.
The best surface roughness for these trials was 5.84μm, obtained in
ZP131 moulds with SC1 magnesium alloy, which was better than
recommended values of 6-13μm.
The optimum percent elongation for these trials was 2.56%, obtained in
the ZP131 moulds with SC1 alloy, and above the recommended value of
2%.
The best UTS for these trials was 170MPa, which was also obtained in the
ZP131 moulds with AZ91 magnesium alloy. This was larger than the
recommended value of 160MPa.
Photomicrographic evidence suggests that the metals produced from static
casting were similar to those produced from traditional foundry practice.
Centrifugal Casting
3D printed moulds were successful in centrifugal casting of both aluminium
and magnesium under varying process conditions
Speeds ranging from 150-450RPM were used without any significant signs
of mould erosion or deformation.
190
The experimental factors considered for CC trials, namely rotational
speed, cavity angle and runner to cavity ratio, together accounted for a
large portion of the total variance in the experimental results.
The tensile properties of centrifugally cast aluminium specimens were
initially below expected levels, due to the presence of hydrogen gas and
subsequent gas porosity, i.e. when the molten metal was not degassed
prior to pouring.
When lance degassing was subsequently used, the resulting UTS
properties were some 10MPa higher than the recommended values.
Microstructural observations then showed smaller grain sizes at high
rotational speeds, and improved tensile properties.
Changes to melt treatment and experimental setup were recommended to
reduce porosity and oxides in the castings.
Overall, the project objectives were achieved:
RP moulds were found to be suitable for casting both aluminium and
magnesium alloys using static and centrifugal filling techniques.
Mechanical properties and microstructural characteristics of castings
varied, with process conditions and the optimum values obtained being
comparable to traditional casting results.
Specifically, magnesium alloys were successfully cast in these moulds,
verifying its potential suitability for rapid casting.
The suitability of the static and centrifugal processes for functional parts to be
rapidly produced with light metals was established. It is postulated that the
future use of pattern-less moulding will gain momentum and that rapid casting
will become standard practice in a variety of situations.
191
REFERENCES
1. Grimm, T., User's guide to rapid prototyping. Illustrated ed. 2004: SME.
404
2. Pham, D.T. and S.S. Dimov, Rapid Manufacturing. 2001: Springer-
Verllag London Limited.
3. Zhang, G., Y.-C. Tsou, and A.L. Rosenberger, Reconstruction of the
Homunculus skull using a combined scanning and stereolithography
process. Rapid Prototyping Journal, 2000. 6(4): p. 267-275.
4. Custom Part. Additive Fabrication. 2010 [cited 2010 28th April];
Available from: http://www.custompartnet.com/wu/additive-fabrication.
5. Ryall, C. and D. Wimpenny. Rapid Prototyping for Rapid Castings.
2008 [cited 2010 April 28th]; Available from:
http://www.jharper.demon.co.uk/rptc01.htm.
6. 3D printing process, in ESG Associates Corporation. 2003.
7. Chua, C.K., K.F. Leong, and C.S. Lim, Rapid prototyping: principles
and applications, ed. C.S. Lim. 2003: World Scientific.
8. Rapid Prototyping, in Foundry Management and Technology. 2004. p.
79-81.
9. Tromans, G., Developments in rapid casting. 2004: John Wiley and
Sons.
10. Pham, D.T. and S.S. Dimov, Rapid prototyping and rapid tooling—the
key enablers for rapid manufacturing. Proceedings of the Institution of
Mechanical Engineers Part C: Journal of Mechanical Engineering
Science, 2003. 217(1): p. 1-23.
11. Dickens, P.M., R. Stangroom, M. Greul, B. Holmer, K.K.B. Hon, R.
Hovtun, R. Neumann, S. Noeken, and D. Wimpenny, Conversion of RP
models to investment castings. Rapid Prototyping Journal, 1995. 4(1):
p. 4-11.
12. Lee, C.W., C.K. Chun, C.M. Cheah, L.H. Tan, and C. Feng, Rapid
investment casting: direct and indirect approaches via fused deposition
modelling. International Journal of Advanced Manufacturing
Technology, 2004. 23(1-2): p. 93-101.
192
13. Sachs, E., M. Cima, and J. Cornie, Three-Dimensional Printing: Rapid
Tooling and Prototypes Directly from a CAD Model. CIRP Annals -
Manufacturing Technology, 1990. 39(1): p. 201-204.
14. Curodeau, A., E. Sachs, and S. Caldarise, Design and Fabrication of
Cast Orthopedic Implants with Freeform Surface Textures from 3-D
Printed Ceramic Shell. Journal of Biomedical Materials Research Part
B: Applied Biomaterials, 2000. 53(5): p. 525-535.
15. Gibbons, D.G. Rapid Casting Using Laser Sintering Sand Moulds and
Cores. Warwick Formula Student- A Case Study 2007 [cited 2010
10th January ]; Available from:
http://www.imeche.org/NR/rdonlyres/1081D434-E678-4961-BDB0-
99AB9877AED9/0/greggibbonssintering.pdf.
16. Tang, Y., J.Y.H. Fuh, Loha, H. T., Y.S. Wong, and L. Lu, Direct laser
sintering of a silica sand Journal of Materials & Design, 2003. 24(8): p.
623-629.
17. Casalino, G., L.A.C. De Filippis, and A. Ludovico, A technical note on
the mechanical and physical characterization of selective laser sintered
sand for rapid casting Journal of Materials Processing Technology,
2005. 166(1): p. 1-8.
18. Kochan, A., Rapid prototyping gains speed, volume and precision.
Assembly Automation, 2000. 20(4): p. 295-299.
19. Waurzyniak, P., Rapid metal, in Manufacturing Engineering. 2003,
ABI/INFORM Global. p. 49-57.
20. Bak, D., Rapid prototyping or rapid production? 3D printing processes
move industry towards the latter. Assembly Automation, 2003. 23(4): p.
340-345.
21. Rebros, M., S.N. Ramrattan, M.K. Joyce, and P.G. Ikonomov,
Behaviour of 3D Printed Sand at Elevated Temperature, in AFS
Transactions, A.F.s. Society, Editor. 2007, American Foundrymen's
Society: Schaumburg. p. 341-348.
22. Bassoli, E., A. Gatto, L. Iuliano, and M.G. Violante, 3D printing
technique applied to rapid casting. Rapid Prototyping Journal, 2007.
13(3): p. 148-155.
193
23. Dimitrov, D., K. Schreve, and N. de Beer, Advances in three
dimensional printing – state of the art and future perspectives. Rapid
Prototyping Journal, 2006a. 12(3): p. 136-147.
24. Dimitrov, D., W. van Wijck, and N. de Beer, Development, evaluation,
and selection of rapid tooling process chains for sand casting of
functional prototypes Journal of Engineering Manufacture, 2007.
221(9): p. 1441-1450.
25. Dimitrov, D., W. van Wijck, K. Schreve, and N. de Beer, Investigating
the achievable accuracy of three dimensional printing Rapid
Prototyping Journal, 2006b. 12(1): p. 42-52.
26. McKenna, N., S. Singamneni, O. Diegel, D. Singh, and A.R.
Choudhury, Rapid manufacture in light metals processing. Materials
Science Forum, 2009. 618-619: p. 387-390.
27. Brown, J.R., Foseco foundryman's handbook. 1994: Butterworth-
Heinemann.
28. Bassoli, E. and E. Atzeni, Direct metal rapid casting: mechanical
optimisation and tolerance calculation. Rapid Prototyping Journal,
2009. 15(4): p. 238–243.
29. Kaplas, M. and R. Singh, Experimental investigations for reducing wall
thickness in zinc shell casting using three-dimensional printing.
Proceedings of the Institution of Mechanical Engineers, Part C: Journal
of Mechanical Engineering Science, 2008. 222(12): p. 2427-2431.
30. Gill, S.S. and M. Kaplas, Comparative Study of 3D Printing
Technologies for Rapid Casting of Aluminium Alloy. Materials and
Manufacturing Processes, 2009. 24(12): p. 1405–1411.
31. Singh, J.P. and R. Singh, Investigations for a statistically controlled
rapid casting solution of lead alloys using three-dimensional printing.
Proceedings of the Institution of Mechanical Engineers, Part C: Journal
of Mechanical Engineering Science, 2009. 223(9): p. 2125-2134.
32. Eliezer, D., E. Aghion, and F.H. Froes, Magnesium Science,
Technology and Applications. Advanced Performance Materials, 1998.
5(3): p. 201-212.
194
33. Eigenfeld, K., W. Tilch, S. Erchov, and O. Podobed, Integrated
magnesium technology. Advanced Engineering Materials, 2004. 6(7):
p. 520-525.
34. Bronfin, B. and E. Aghion, Magnesium Alloys Development towards the
21st Century. Materials Science Forum, 2000. 350-351: p. 19-30.
35. Hanawalt, J.D. and M. Okada, Oxidation inhibiting techniques for
magnesium sand casting, in Transactions of the American
Foundrymen's Society, T.o.t.A.F.s. Society, Editor. 1972, Transactions
of the American Foundrymen's Society. p. 87-90.
36. Fantetti, N., A. Thorvaldsen, and A. Couture, Properties of Magnesium
Plaster Casting. SAE Technical Paper Series, 1991: p. 1-10.
37. Lun Sin, S., D. Dube, and R. Tremblay, Interfacial reactions between
AZ91D magnesium alloy and plaster mould material during investment
casting. Materials science and technology, 2006. 22(12): p. 1456-1463.
38. Idris, M.H. and A.J. Clegg, Processing and evaluation of investment
cast magnesium-base alloy, in Transactions of the American
Foundrymen's Society. 1996, Transactions of the American Foundry's
Society. p. 237-244.
39. Takamori, S., Y. Osawa, T. Kimura, L. Xinbao, and T. Muka, Casting
Surface of AZ91 Alloy and Its Reaction with Sand Mold. Materials
transactions, 2008. 49(5): p. 1089-1092.
40. Cingi, C., J. Vainola, and j. Orkas, Role of oxygen in mold metal
reactions in investment casting of magnesium alloy AZ91E.
International Foundry Research, 2007. 59: p. 18-27.
41. Kim, S., M. Kim , T. Hong, H. Kim, and Y. Kim, Investment casting of
AZ91HP magnesium alloy Metals and Materials International, 2000.
6(3): p. 275-279.
42. Gjestland, H. and H. Westengen, Advancements in high pressure die
casting of magnesium. Advanced engineering materials 2007. 9(9): p.
769-776.
43. Strieter, F.P. and R.J. Maonner, Centrifugal casting of magnesium part,
in American Foundryman, A.F. Association, Editor. 1946, American
Foundryman Association: Chicago. p. 43-47.
195
44. Chirita, G., D. Soaresa, and F.S. Silva, Advantages of the centrifugal
casting technique for the production of structural components with Al–
Si alloys. Materials & Design, 2008. 29(1): p. 20-27.
45. Mukunda, P.G., A. Shailesh Rao, and S.S. Rao, Influence of rotational
speed during centrifugal casting on sliding wear behaviour of the Al-2Si
alloy. Frontiers of Materials Science in China, 2009. 3(3): p. 339-344.
46. Howson, H.O., Macrostructural comparison of centrifugal and static
castings. Foundry Trade Journal, 1969. 127(2750): p. 261-272.
47. Mesquita, R.A., D.R. Leiva, A.R. Yavari, and W.J. Botta Filho,
Microstructures and mechanical properties of bulk AlFeNd(Cu, Si)
alloys obtained through centrifugal force casting. Materials Science and
Engineering A, 2007. 452-453: p. 161-169.
48. Meinecke, G., Production, properties and processing directions for
moulding and casting plasters. Zement-kalk-Gips, Edition B, 1989.
42(5): p. 233-235.
49. Corporation, Z. (2007) ZB60 MSDS. 1-9.
50. Corporation, Z. (2002) ZB 58 Binder MSDS. 1-8.
51. Cochran, W.G. and G.M. Cox, Experimental Designs. 1950: Wiley.
52. Rekab, K., Statistical Design of Experiments with Engineering
Applications. 2005: CRC Press.
53. Beeley, P.R., Foundry technology. 2 ed. 2001: Butterworth-Heinemann.
719 pages.
54. ASTM International, D7012 – 10 Standard Test Method for
Compressive Strength and Elastic Moduli of Intact Rock Core
Specimens under Varying States of Stress and Temperatures. 2010. p.
9.
55. Brown, J.R., Foseco Non-Ferrous Foundryman's Handbook. 1999:
Butterworth-Heinemann.
56. Bettles, C.J., C.T. Forwood, D.S. Jones, J.R. Griffiths, M.T. Frost , D.H.
St.John, Ma Qian, G.-L. Song, and J.F. Nie, AMC-SC1: A New
Magnesium Alloy Suitable for Powertrain Applications SAE
International Technical Papers, 2003.
57. American Foundrymen's Society, Aluminum Casting Technology 2nd
edition, ed. 2nd. 1993: American Foundrymen's Society. 368.
196
58. Roy, R., Primer on the taguchi method. 2010: SME.
59. Vander Voort, G.F., Color Metallography, in MICROSCOPY TODAY.
2005. p. 22-27.
60. ASM Handbook Committee, ASM Handbook. Metallography and
Microstructures, ed. G.F. Vander Voort. Vol. 9. 2004: ASM International
1184.
61. ASTM International, E 112 – 96 Standard Test Methods for
Determining Average Grain Size. 2004, ASTM International.
62. Rennie, A.E.W., C.E. Bocking, and D.M. Jacobson, Rapid and Virtual
Prototyping and Applications. 2003: John Wiley and Sons.
63. Smallman, R.E. and R.J. Bishop, Modern physical metallurgy and
materials engineering: science, process, applications. 6 ed. 1999:
Butterworth-Heinemann. 438.
64. Aluminum Alloy Castings. 2004: ASM International.
65. Hong, S.J., S.S. Kim, J.H. Lee, Y.N. Kwon, Y.S. Lee, and J.H. Lee,
Effect of microstructural variables on tensile behaviour of A356 cast
aluminium alloy. Materials Science and Technology, 2007. 23(7): p.
810-814.
66. Kobayashi, T., Strength and fracture of aluminum alloys. Materials
Science and Engineering A, 2000. 286(2): p. 333-341.
67. Lee, C.D., Effect of grain size on the tensile properties of magnesium
alloy. Materials Science and Engineering A, 2007. 459(1-2): p. 355–
360.
68. Campbell, J., Castings: The New Metallurgy of Cast Metals. 2003:
Butterworth-Heinemann.
69. Backerud, C., G. Chai, and J. Tammen, Solidification Characteristics of
Aluminium Alloys. 1990: AFS International.
70. St John, D.H., A.K. Dahle, T. Abbott, M.D. Nave, and Qian Ma,
Solidification of cast Magnesium alloys. TMS Annual Meeting, 2003.
71. Avedesian, M.M., H. Baker, and ASM International Handbook
Committee, Magnesium and magnesium alloys. 2nd ed. 1999: ASM
International. 314.
197
72. Dunlop, G., Bettles, C.J., J.R. Griffiths, K. Venkatesan, L. Zheng, and
M. Qian. The Effect of Grain Size on the Mechanical Properties of
AMSC1. in 6th International Conference of Magnesium Alloys and
Their Applications. 2003. Wolfsburg.
73. Standards, B., Specification for Aluminium and Aluminium alloy ingots
and castings for general engineering purposes. 1988, British
Standards.
74. Bettles, C.J., M.A. Gibson, and Z. S.M., Microstructure and mechanical
behaviour of an elevated temperature Mg-rare earth based alloy.
Materials Science and Engineering A, 2009. 505(1-2): p. 1-12.
75. Lu, Y.Z., Q.D. Wang, W.J. Ding, X.Q. Zeng, and Y.P. Zhu, Fracture
behaviour of AZ91 magnesium alloy. Materials Letters, 2000. 44: p.
265-268.
76. Ceschini, L., A. Morri, A. Morri, Gamberini, A. , and S. Messieri,
Correlation between ultimate tensile strength and solidification
microstructure for the sand cast A357 aluminium alloy. Materials and
Design, 2009. 30(10): p. 4525–4531.
77. Totten, G.E. and D.S. Mackenzie, Handbook of Aluminum. Vol. 1.
2003: CRC Press. 1310 pages.
78. Committee, A.I.H., Alloy Phase Diagrams. ASM Handbook, ed. H.
Baker and H. Okamoto. Vol. 3. 1992: ASM International.
79. Nave, M.D., A.K. Dahle, and D.H. St John, Eutectic growth
morphologies in Magnesium-Aluminium alloys, in Magnesium
technology 2000. 2000, TMS. p. 233-242.
80. Choong Do Lee and Kwang Seon Shin, Effect of microporosity on the
tensile properties of AZ91 magnesium alloy. Acta Materialia, 2007.
55(13): p. 4293–4303.
81. Zeng, X., Y. Wang, Wenjiang Ding, A.A. Luo, and A.K. Sachdev, Effect
of Strontium on the Microstructure, Mechanical Properties, and
Fracture Behaviour of AZ31 Magnesium Alloy. Metallurgical and
materials transactions A, 2006. 37A: p. 1333-1341.
82. Polmear, I., Light Alloys. 2005: Butterworth-Heinemann. 437.
198
83. Davis, J.R., Davis J. R. & Associates, and ASM International.
Handbook Committee, Aluminum and aluminum alloys - ASM specialty
handbook. 5, illustrated ed, ed. J.R. Davis. 1993: ASM International.
784 pages.
84. Fatahalla, N., M. Hafiz, and M. Abdulkhalek, Effect of microstructure on
the mechanical properties and fracture of commercial hypoeutectic Al-
Si alloy modified with Na, Sb and Sr. Journal of Materials Science,
1999. 34: p. 3555 – 3564.
85. Hegde, S. and K.N. Prabhu, Modification of eutectic silicon in Al–Si
alloys. Journal of Materials Science, 2008. 43(9): p. 3009–3027.
86. Shabestari, S.G. and F. Shahri, Influence of modification, solidification
conditions and heat treatment on the microstructure and mechanical
properties of A356 aluminum alloy. Journal of Materials Science, 2004.
39(6): p. 2023 – 2032.
87. Haque, M.M. and A.F. Ismail, Effect of superheating temperatures on
microstructure and properties of strontium modified aluminium–silicon
eutectic alloy. Journal of Materials Processing Technology, 2005. 162–
163: p. 312–316.
88. Wang, Q.G., Microstructural Effects on the Tensile and Fracture
Behaviour of Aluminum Casting Alloys A356/357. Metallurgical and
Materials Transactions A, 2007. 34(12): p. 2887-2899.
89. Ma, Z., A.M. Samuel, F.H. Samuel, H.W. Doty, and S. Valtierra, A
study of tensile properties in Al–Si–Cu and Al–Si–Mg alloys: Effect of
β-iron intermetallics and porosity. Materials Science and Engineering
A, 2008. 490: p. 36–51.
90. Taylor, J.A., The Effect of Iron in Al-Si Casting Alloys in 35th Australian
Foundry Institute National Conference, C.e. al, Editor. 2004, Australian
Foundry Institute (AFI): Adelaide, South Australia. p. 148-157.
91. Bourcier, R.J., D.A. Koss, R.E. Smelser, and O. Richmond, The
influence of porosity on the deformation and fracture of alloys. Acta
Metallurgica, 1986. 34(21): p. 2443-2453.
199
92. Ammar, H.R., A.M. Samuel, and F.H. Samuel, Porosity and the fatigue
behaviour of hypoeutectic and hypereutectic aluminum–silicon casting
alloys. International Journal of Fatigue, 2008. 30(6): p. 1024–1035.
93. ASM Handbook Committee, ASM Handbook. Fractography, ed. K.
Mills. Vol. 12. 2004: ASM International 1184.
94. Janco, N., Centrifugal casting. 1988: American Foundrymen's Society.
95. Johnson, H.V., Manufacturing Process. 2 ed. 1984: DIANE Publishing.
590
96. Subramanya, K., 1000 solved problems in fluid mechanics (includes
hydraulic machines)2005: Tata McGraw-Hill. 456 pages.
97. Chirita, G., I. Stefanescu, D. Soares, and F.S. Silva, Centrifugal versus
gravity casting techniques over mechanical properties
Anales de Mecánica de la Fractura, 2006. 1: p. 317-322.
98. Lu, L. and A.K. Dahle, Iron-Rich Intermetallic Phases and Their Role in
Casting Defect Formation in Hypoeutectic Al-Si Alloys. Metallurgical
and Materials Transactions A, 2005. 36A: p. 819-835.
99. Stefanescu, D.M., ASM Handbook. Vol. 15 Casting. 1988.
100. Monroe, R., Porosity in Castings. Transactions of the American
Foundry Society, 2005. 113(Silver Anniversary Paper): p. 519 - 546.
101. Choong Do Lee, Effects of microporosity on tensile properties of A356
aluminium alloy. Materials Science and Engineering: A, 2007. 464(1-2):
p. 249-254.
102. Herrera, A. and V. Kondic. in Proceedings of the International
Conference on Solidification and Cast Metals. 1977. Sheffield, UK.
103. Campbell, J., Castings. 2 ed. 2003: Butterworth-Heinemann. 335
pages.
104. Campbell, J., Castings practice: the 10 rules of castings. 2004:
Butterworth-Heinemann.
105. Campbell, J., An Overview of the Effects of Bifilms on the Structure and
Properties of Cast Alloys. Metallurgical and Materials Transactions B,
2005. 37B: p. 857-863.
200
106. John, K., P.R. Beeley, J.R. Griffiths, and N.R. Green, Commentaries on
'Entrainment defects' by J. Campbell. Materials Science and
Technology, 2006. 22(8): p. 999-1008.
107. De-Feng, M., H. Guo-Qiu, H. Zheng-Fei, Z. Zheng-Yu, C. Cheng-Shu,
and Z. Wei-Hua, Crack initiation and propagation of cast A356
aluminum alloy under multi-axial cyclic loadings. International Journal
of Fatigue, 2008. 30(10-11): p. 1843-1850.
108. British Standard, BS EN 10 002-1 Tensile testing of metallic materials,
in Part 1. Method of test at ambient temperature 1990, British
Standards Institution p. 1-17.
A-1
APPENDIX A Mould Material
A.1 Experimental Design
Firstly consider a square matrix, n x n, shown in equation (A.1):
x { } [y] (A.1)
Where:
[x] = X (design) matrix
{ }= Unknown constants (column matrix)
{ }y = Experimental results
So in the square matrix case, shown in equation (A.1), above, the unknowns
are the missing constants in the column matrix, β But this assumes that the
missing column matrix is square which in this case it is not, as it contains 6
missing constants in one column i.e. 6x1 matrix.
To allow further analysis to be done the column matrix must be multiplied by
its transposed matrix, shown in equation (A.2)
If, 1 2 3X a a a then its transpose;
1
T
2
3
a
X a
a (A.2)
Thus:
T T[X][X ]{ } [Y][X ] (A.3)
Rearranging equation (A.3) β we obtain equation (A.4):
A-2
T T 1{ } ([Y][X ]) ([X][X ]) (A.4)
Where:
0
1
2
11
22
12
{ }
(A.5)
The missing column matrix in equation (A.5) shows the constants that will
form the polynomial expression, shown in equation (A.6) below.
2 2
0 1 1 2 2 11 1 22 2 12 1 2y X X X X X X (A.6)
The missing column matrix was determined with the use of Matlab and Mintab
software. The procedure was to establish the values missing co-efficients,
based off the coded values, in Matlab then confirming these in Minitab. Also
Minitab was used to confirm the values of the natural variables, obtained by
hand calculations. A sample of these calculations is shown below.
Co-efficient values found from the coded variables Matlab for ZP131
permeability strength model were:
2099.08
-109.377
-105.208{ }
82.8813
166.930
40.5881 (A.7)
Thus, the expression in terms of natural variables becomes:
A-3
1 2 1 2
2 2
1 2 1 2
P X ,X 2099.08 109.33X 105.208X
82.8813X 166.93X 40.5881X X (A.8)
To convert a coded variable into a natural variable, the increment of change of
the natural variable and the centre point must be known, i.e.:
N Cx
i (A.9)
Where:
x = coded variable
N = Value of natural variable
C = Centre point of natural variable in experimental design
i = Increment of change of natural variable
The configuration of this formula is shown below for temperature and time of
mould baking
1
T 200X 0.02T 4
50 (A.10)
and 2
t 6X 0.5t 3
2 (A.11)
Equations (A.10) and (A.11) are then substituted into equation (A.8), which
gives the following:
2 2
P T,t 2099.08 109.33 0.02T 4 105.208 0.5t 3
82.8813 0.02T 4 166.93 0.5t 3
40.5881 0.02T 4 0.5t 3 (A.12)
Expanding equation (A.12)
A-4
2 2
MP T,t 2099.08 2.1876T 437.52 29.89t 179.32
0.039408T 15.7532T 1576.32 29.68t 356.16t
1068.48 0.4059Tt 2.435T 81.18t 487.08 (A.13)
Collecting like terms in equation (A.13) gives the final form of the model in
natural variables
2 2
MP T,t 5847.32 20.3758T 467.23t
0.039408T 29.68t 0.4059Tt (A.14)
The co-efficients in terms of the natural variables are then shown below in
equation (A.14) in matrix form. These values were checked against those
obtained in Minitab, which confirms that these values were correct.
5847.82
20.3858
467.23{ }
0.039408
39.68
0.4059 (A.15)
A.2 Permeability and Compressive Stress Calculations
Permeability
The permeability of the tested mould samples was calculated using Darcy‟s
equation, shown below.
mean in,out
Q PLk
AP P 1000 (A.16)
A-5
Where:
= Dynamic viscosity of passing fluid air (1.82 x 10-5 Ns/m2)
L = length of specimen (m)
A = Cross sectional area of specimen (m2)
Pmena = Mean pressure of inlet and outlet added onto Patm
in,outP = Inlet pressure – outlet pressure
Q = Flow rate m3/s
P = Pressure difference between atmosphere and gauge pressure
Compressive strength
The compressive strength values were calculated by using the definition of
stress, shown below in equation (A.17). The ultimate compressive strength
was determined when the specimen had started to crack and the force either
reduced or remained relatively constant.
F
A (A.17)
Where:
= The resulting compressive stress
F = The applied uniaxial force
A = The cross-sectional area, perpendicular to the applied force
A.3 Raw Experimental Results From Material Testing
Over the next few pages the raw experimental data is presented relating to
the permeability and compressive testing conducted on both the ZP131 and
ZCast501 mould materials.
A-6
A.3.1 Permeability Data
Table A.1 ZP131 mould material permeability data
Temp
(°C)
Time
(Hrs) Outflow rate (m3/s) Inlet pressure (KPa) Mean Press (KPa) K(m2) K (mD)
150 4 3.48E-05 16 110.3 2.5E-12 2496
250 4 2.92E-05 16 110.3 2.1E-12 2092
150 8 2.97E-05 16 109.2 2.1E-12 2150
250 8 2.66E-05 16 110.3 1.9E-12 1909
130 6 3.53E-05 16 110.3 2.5E-12 2533
271 6 3.22E-05 15.5 110.3 2.3E-12 2371
200 3.17 3.48E-05 16 110.6 2.4E-12 2473
200 8.23 3.53E-05 16 110.6 2.5E-12 2509
200 6 3.48E-05 15.6 110.4 2.5E-12 2541
200 6 2.86E-05 16 109.2 2.0E-12 2062
200 6 2.81E-05 16 109.2 2.0E-12 2025
200 6 2.35E-05 16 109.2 1.7E-12 1694
200 6 3.02E-05 16 109.2 2.1E-12 2173
A-7
Table A.2 Experimental permeability data for ZCast501 mould material
Temp
(°C)
Time
(Hrs) Outflow rate (m3/s) Inlet pressure (KPa) Mean Press (KPa) K(m2) K (mD)
150 4 1.63E-05 12.5664615 108.6 1.5E-12 1500
250 4 2.05E-05 12.3406667 108.4 1.9E-12 1917
150 8 2.17E-05 10.9258043 106.7 2.3E-12 2304
250 8 1.89E-05 12.03 108.3 1.8E-12 1825
130 6 1.60E-05 9.87268182 107.2 2.1E-12 2120
271 6 1.79E-05 9.47026316 107.3 2.2E-12 2213
200 3.17 1.84E-05 14.9532263 110.0 1.4E-12 1405
200 8.23 1.69E-05 10.2588692 107.7 1.9E-12 1926
200 6 1.87E-05 12.4919192 108.8 1.7E-12 1721
200 6 2.46E-05 11.6226814 107.0 2.5E-12 2499
200 6 2.10E-05 12.5545169 107.5 1.9E-12 1948
200 6 2.30E-05 12.0973137 107.2 2.2E-12 2241
200 6 2.20E-05 11.7914585 107.1 2.1E-12 2175
A-8
A.3.2 Compressive Strength Data
Table A.3 ZP131 and ZCast501 compressive strength data
Coded units Un-coded units ZP131 Compressive strength ZCast501 Compressive strength
X1 X2 Temp (°C) Time (Hrs) Force (N) UCS (MPa) Force (N) UCS (MPa)
-1 -1 150 4 871.33 1.78 1674 0.85
1 -1 250 4 401.66 0.82 988 0.50
-1 1 150 8 896.33 1.83 1803 0.92
1 1 250 8 248.33 0.51 515.30 0.26
-1.414 0 130 6 837.66 1.71 1938 0.99
1.414 0 270 6 357.33 0.73 755 0.38
0 -1.414 200 3.17 590.66 1.20 2031.30 1.03
0 1.414 200 8.88 640.33 1.30 561.70 0.29
0 0 200 6 424 0.86 2056 1.05
0 0 200 6 572.33 1.17 2431 1.24
0 0 200 6 738.33 1.50 1744 0.89
0 0 200 6 590.33 1.20
0 0 200 6 778.33 1.59
B-1
APPENDIX B STATIC CASTING INFORMATION
B.1 Design of Experiments and ANOVA
The L9 experimental design used in the static casting trials, is able to identify
the highest ranking factor and level combinations, however, an ANOVA is also
needed on the signal to noise ratio, to establish at what confidence level are
these factors significant. The AVOVA table is used here to identify any
significant factors at 90, 95 or 99 % confidence level. Over the following
section, the individual calculations presented in the ANOVA table are
presented in the context of the L9 experimental design.
Degrees of freedom
Factor D.O.F:
D.O.F # 1factor factor levels (B.1)
For the L9 experiment each factor is at three levels so the D.O.F associated
with factor is = 3-1 = 2 D.O.F.
Total D.O.F:
totalD.O.F # exprimental trials 1 (B.2)
For the L9 experiment there were 9 trials so D.O.F = 9-1 = 8
Error D.O.F:
FactorD.O.F D.O.F D.O.FError total (B.3)
B-2
I.e. for a L9 experiment = 8-(number of factors) *(2)
In this case we have 4 factors at 2 levels thus D.O.F 8 2 2 2 2 0Error
As we can see we now have used all the columns in the experiment and
consequently have zero D.O.F for the error in the experiment. To account for
some error the lowest factor is normally pooled according to some criteria i.e.
least significant factor is assumed as the error to allow further ANOVA
calculations (as they require a known amount of error).
B.1.1 Sum of squares
Each factor is considered a column in the Taguchi design. The formula below
returns the value of the sum of square of each factor based on its S/N ratios.
1 2 31 2 3
2 2 2
f f fFactor f f fSS K T T K T T K T T
(B.4)
Where:
1...nfT = Average of response of factor f1
T = Average response
1...nfK = Number of levels of factor K
Variance
Factor Variance: . .
factor
factor
SSv
DO F (B.5)
E.g. Variance of mould coating = 3.02/2 = 1.51
B-3
F-Ratio
The calculated F value is a simple ratio of variances. When this ratio becomes
large enough the two sample variances are accepted as being unequal at
some confidence level. We use F-ratio to see if our value is significant or
greater than the standard table value at a specific confidence level usually 90,
95 or 99 %. If the experimental value is higher than the standard value at the
specific confidence interval (taking into account the D.O.F) the factor is said to
be significant and that the variation in the results cannot be explained by
chance alone (based on the F distribution).
factorratio
error
vF
v (B.6)
The critical F value, Fcritical is found from standard tables at the required
confidence level and at the appropriate D.O.F.
A factor is said to be significant when the Fratio is larger than the Fcritical value.
Percent contribution
Percent contribution is simply the ratio of the sum of squares of each factor to
the total sum of squares i.e. how much variation does each factor contribute
to the total variance.
factorfactor
total
SS%
SS (B.7)
Pooling of factors
The following is taken from „Taken from A primer on the Taguchi method by
Ranjit K. Roy [58]. Error terms in an experiment represent the degree of inter-
B-4
experiment error when the D.O.F is sufficiently large. When the error D.O.F is
small or zero which is the case when all the columns (factors) in an
orthogonal array are occupied, the small column effects are pooled to form a
larger error term (known as pooling up). In terms of what has been done in the
current experiment, the factor contributing the least sum of squares to the
total sum of squares was pooled as the error. It is mentioned that repetition of
result can provide an error term even if the experiment has all its columns
used.
B.2 Mechanical testing data
Table B.1 Results of tensile testing of static castings (average values shown in bold).
Testing was conducted at 1mm/min.
Trail
number
Gauge
dia. (mm)
Gauge
Length
Extension
(m)
Force at
break
UTS
(MPa)
Ext
(%)
1A 0.0048 0.05 0.001325 3087.667 170.63 2.65
1B 0.00496 0.047 0.00111 3064.333 158.59 2.36
1C 0.00492 0.05 0.0009 2724 143.28 1.80
1D 0.005 0.05 0.000935 3007.33 153.16 1.87
156.42 2.17
2A 0.00496 0.05 0.00132 2676.333 138.51 2.64
2B 0.00508 0.0479 0.00104 2582.667 127.42 2.17
2C 0.00475 0.0505 0.00129 2221 125.33 2.55
2D 0.0048 0.05 0.001438 2330.667 128.80 2.88
130.02 2.56
3A 0.00416 0.05 0.00048 1906 140.23 0.96
3B 0.0041 0.0549 0.000375 1466.33 111.06 0.68
3C 0.00414 0.049 0.0003 1561.33 115.99 0.61
3D 0.00374 0.05 0.00037 1551.667 141.24 0.74
127.13 0.75
4A 0.00454 0.048 0.001229 2482 153.32 2.56
B-5
4B 0.00417 0.048 0.001116 1798.667 131.70 2.33
4C 0.00503 0.0485 0.00109 2245.333 112.99 2.25
4D 0.00493 0.05 0.001 2175.33 113.96 2.00
127.99 2.28
5A 0.00498 0.0505 0.000375 2616 134.30 0.74
5B 0.00506 0.0505 0.00046 2478 123.23 0.91
5D 0.00515 0.049 0.00044 2887.333 138.61 0.90
132.05 0.85
6B 0.00367 0.04935 0.00075 1596 150.87 1.52
150.87 1.52
7A 0.00492 0.0491 0.000285 1514.33 79.65 0.58
7B 0.00488 0.0504 0.00045 1912.667 102.26 0.89
7C 0.00492 0.0491 0.00035 1861.33 97.90 0.71
93.27 0.73
8A 0.0051 0.049 0.0006 2105.33 103.06 1.22
8B 0.00504 0.0477 0.0009 2816.333 141.17 1.89
8C 0.00508 0.0468 0.00071 2778.667 137.09 1.52
127.11 1.54
9A 0.00475 0.049 0.001505 2730 154.06 3.07
9B 0.00453 0.0484 0.0019 2655 164.73 3.93
9C 0.00508 0.0505 0.00155 3108.667 153.38 3.07
157.39 3.36
Table B.2 Brinell hardness results, conducted on cast samples. (All samples were
tested at a minimum of 4 readings. Samples were tested with 500Kg load with 10mm
diameter ball, with readings 1 and 2 showing the measured diameters of the
indentation)
Sample Read
1(mm)
Read
2
Ave Read
1
Read
2
Ave Total
Ave
HB
1A 3.4 3.3 3.35 3.3 3.2 3.25 3.3 56.82
1B 3.4 3.2 3.3 3.4 3.2 3.3
2 3.5 3.6 3.55 3.5 3.7 3.6 3.575 48.17
B-6
3 3 3.2 3.1 3.3 3.3 3.3 3.2 60.54
4 3.6 3.5 3.55 3.6 3.5 3.55 3.55 48.87
5 3.1 3.1 3.1 3.1 3 3.05 3.075 65.70
6 3.2 3.3 3.25 3.1 3.2 3.15 3.2 60.54
7 3.4 3.3 3.35 3.5 3.5 3.5 3.425 52.63
8 3.3 3.4 3.35 3.1 3.4 3.25 3.3 56.82
9 3.5 3.7 3.6 3.5 3.7 3.6 3.6 47.46
Table B.3 Surface roughness values measured from the as-cast surface.
Trial Surface roughness readings, Ra (μm) Overall
average
1 9.2561 11.5939 7.6289 8.4369 10.4254 9.6500 9.4985
2 3.7006 10.4743 7.4631 9.3854 5.6562 7.2831 7.3271
3 7.1392 6.2032 5.0519 4.2915 4.9427 7.4167 5.8409
4 38.1190 32.7692 20.0400 7.7737 17.9628 25.3097 23.6624
5 15.4813 8.2465 10.7501 14.0728 6.9355 13.3186 11.4675
6 13.0575 22.6259 11.6315 10.0185 12.8754 15.7890 14.3330
7 17.7029 7.9240 11.2059 13.6704 11.4595 13.8245 12.6312
8 11.6313 13.7565 16.5517 14.4340 13.4570 14.0485 13.9798
9 20.0345 9.2689 11.2082 12.7540 9.1077 8.3472 11.7868
B.2.1 ASTM grain size
To determine the average dendrite grain size of the static castings, the
Planimetric method outlined in the ASTM grain counting standard, E 112 – 96
[61] was used. In compliance with this standard, five different fields were
considered in each trial. To determine the average grain size, the number of
grains in a given area must be counted. For this analysis, a magnification of
100 was chosen and the area of the photomicrograph was computed by a
standard measurement device.
The actual counting procedure considers a whole grain inside the area as
equal to one with grains overlapping at the edge of the picture considered as
B-7
a half. From square photographs, the four corners are considered as one
whole grain. Once all the five samples for each field was counted the average
number of grains was calculated. From here the value of Na was calculated
with equation (B.8).
int ercepted
inside
NNa f(N 1)
2 (B.8)
Where: f2
2
magification
Area(mm )
NInside = Number of grains inside given area
NInterepted = Number of grains intercepted at boundary area
Area = Area of photomicrograph (allowing for magnification)
In these trials magnifications of X100 was used. The size of the
photomicrograph was 70mmX53mm. Once Na is determined, the ASTM
average grain size is found by the following equation:
10G (3.32log Na) 2.954 (B.9)
Once G has been determined, the average grain properties such as area and
diameter can be found from standard tables correlating the ASTM grain size.
B-8
Table B.4 Experimental results from grain size testing
Trial Field 1 Field 2 Field 3 Field 4 Field 5 Average G
1 7 10 3 4 8 6.4 1.16
2 19 14 32 29 39 26.6 3.21
3 15.5 19 17 19 23 18.7 2.71
4 30 23 22 21 25 24.2 3.08
5 15.5 17 16.5 17.5 18 16.9 2.56
6 8 3 9 9 6 7 1.29
7 15.5 16 15.5 17.5 13 15.5 2.43
8 18 30 28 21 25 24.4 3.09
9 67 46 75 92 71 70.2 4.61
B-9
B.3 Microstructure Evaluation
(X100) (X400)
(a) (b)
AZ91 castings produced in ZP131 plaster moulds at 690°C. Mould was coated with
Isomol 200 coating (σUTS = 156.4MPa δ = 2.17% HB = 56.8 Grain size: 1.11)
(c) (d)
AZ91 casting produced in ZCast501 mould cast at 740°C. The mould was coated with
Magcoat. (σUTS = 150.87 MPa δ =1.51% HB = 60.53 Grain size: 1.29)
(e) (f)
AZ91 casting produced in Silica sand foundry mould, cast at 770°C. The mould coating
was Zircoat (σUTS = 127.1MPa δ = 1.54% HB = 56.82 Grain size: 3.09)
Figure B.1 AZ91 photomicrographs of (a&b) ZP131 (c&d) ZCast and (e&f) Silica moulds
B-10
(a) (b)
SC1 magnesium alloy cast in ZP131 plaster moulds at 730°C. The mould was coated
with Zircoat (σUTS = 130.02 MPa δ = 2.56% HB = 48.87 Grain size: 3.08)
(c) (d) SC1 casting poured into ZCast501 moulds at 770°C. The mould was coated with Isomol
200 (σUTS = 128 MPa δ =2.28% HB = 48.16 Grain size: 3.21)
(e) (f)
SC1 casting produced in Silica foundry sand moulds at 690°C. The mould was coated
was Magcoat (σUTS = 157.38MPa δ = 3.35 HB = 47.47 Grain size: 4.6)
Figure B.2 SC1 photomicrographs of (a&b) ZP131 (c&d) ZCast and (e&f) Silica moulds
B-11
(a) (b)
A356 Aluminium castings produced in ZP131 plaster moulds, poured at 730°C. The
mould coating used was Zircoat (σUTS = 140.23MPa δ = 0.96% HB = 60.53 Grain size:
2.45)
(c) (d)
A356 castings produced in ZCast moulds poured at 770°C. The mould coating was
Isomol 200 (σUTS = 134.30MPa δ = 0.74% HB = 66.70 Grain size: 2.45)
(e) (f)
A356 castings produced in Silica foundry sand moulds, poured at 690°C. The mould
coating used was Magcoat (σUTS = 102.26MPa δ = 0.89% HB = 52.65 Grain size: 2.18)
Figure B.3 Al microstructures in (a&b) ZP131 (c&d) ZCast and (e&f) Silica mould
B-12
B.4 Castings Produced in ZP131 Moulds
AZ91
(a) (b)
(c) (d)
Figure B.4 SEM photographs of AZ91 castings produced at (a) X57 (b) X250 (c) X500
and (d) X4000. (Sample properties UTS = 153.16MPa, δ = 1.87%).
B-13
SC1
(a) (b)
(c) (d)
Figure B.5 SEM photos of AM-SC1 alloy at (a) X58 (b) X250 (c) X500 and (d) X4000.
(Sample properties UTS = 127.42MPa, δ = 2.17%).
B-14
A356
(a) (b)
(c) (d)
Figure B.6 SEM photographs of A356 castings at (a) X57 (b) X250 (c) X500 and (d)
X4000. (Sample properties UTS = 141.24MPa, δ = 0.74%).
B-15
B.4.1 Castings Produced in ZCast Moulds
AZ91
(a) (b)
(c) (d)
Figure B.7 AZ91 castings (a) X57 (b) X250 (c) X500 and (d) X4000. (Sample properties
UTS = 150.87MPa, δ = 1.52%).
B-16
SC1
(a) (b)
(c) (d)
Figure B.8 AMSC1 fractured surfaces at (a) X57 (b) X250 (c) X500 and (d) X4000.
(Sample properties UTS = 133.96MPa, δ = 2.00%).
B-17
A356
(a) (b)
(c) (d)
Figure B.9 SEM photos of A356 castings at (a) X57 (b) X250 (c) X500 and (d) X4000.
(Sample properties UTS = 138.61MPa, δ = 0.90%).
B-18
B.5 Tensile Part Setup
Figure B.10 Cad drawing of the tensile test part used in the static casting trials
B.6 Static Mould Design
The static mould design consisted of the pattern, detailed below in Figure
B.11. This cavity was placed in the mould, shown in Figure B.12.
Figure B.11 CAD drawing, showing critical dimensions of mould cavity (pattern) used
in the static mould design.
B-19
Figure B.12 – Configuration of the cavity in the mould.
B.7 Alloy Constitutes
AZ91HP:
%Al %Zn %Mn %Be %Mg
8.70 0.80 0.25 0.001 90.25
SC1:
Nd Ce Zn La Zr Mn Pr
1.69 0.71 0.50 0.40 0.38 0.13 0.07
Fe Ni Cu Si Mg
<0.004 <0.001 <0.002 <0.01 96.10
Al CC601
Si Fe Cu Mn Mg
6.5–7.5 0.55 0.2 0.35 0.2–0.65
C-1
APPENDIX C CENTRIFUGAL CASTING DATA
C.1 Experimental Design
The Box-Behnken experimental design analysis is very similar to the analysis
covered in section A.1. The analysis is identical with the exception that the
missing column matrix has some additional terms, associated with the third
independent variable. The relevant missing column matrix is shown below in.
Where:
0
1
2
3
11
22
33
12
13
23
{ }
(C.1)
These missing co-efficients were then substituted in the second order
polynomial expression shown below in equation (C.2).
2 2
0 1 1 2 2 3 3 11 1 22 2
2
33 2 12 1 2 13 1 3 23 2 3
y X X X X X
X X X X X X X
(C.2)
For completeness, the entire X matrix, is shown below, in coded format
C-2
Table C.1 X matrix used in Box-Behnken experimental design
2 2 2
0 1 2 3 1 2 3 1 2 1 3 2 3
1 1 1 0 1 1 0 1 0 0
1 1 1 0 1 1 0 1 0 0
1 1 1 0 1 1 0 1 0 0
1 1 1 0 1 1 0 1 0 0
1 1 0 1 1 0 1 0 1 0
1 1 0 1 1 0 1 0 1 0
1 1 0 1 1 0 1 0 1 0
1 1 0 1 1 0 1 0 1 0
1 0 1 1 0 1 1 0 0 1
1 0 1 1 0 1 1 0 0 1
1 0 1 1 0 1 1 0 0 1
1 0 1 1 0 1 1 0 0 1
1 0 0 0 0 0 0 0 0 0
1 0 0 0 0 0 0 0 0 0
1 0 0 0 0 0 0 0 0 0
x x x x x x x x x x x x x
x
C-3
C.2 Mechanical Testing Data
Over the following section the experimental results from the mechanical
testing conducted on the centrifugal castings is presented. These results
included, Ultimate Tensile Strength (UTS), Yield Strength, percent elongation
to fracture and grain counting conducted on the microstructures, cut from the
tensile test parts. In the UTS and Yield strength testing, two samples at each
condition were tested, denoted by A and B respectively.
Table C.2 UTS results from centrifugally cast test bars
Trial UTS
(MPa)
Average Trial UTS
(MPa)
Average
1A 90.5 8B 110.3 110.3
1B 94.7 92.6 9A 104.5
2A 131.4 9B 100.9 102.7
2B 125.8 128.6 10A 105.5
3A 104.9 10B 106.5 106
3B 94.3 99.6 11A 116.6
4A 107.8 11B 116.8 116.7
4B 111.3 109.55 12A 109.3
5A 108.9 12B 115.7 112.5
5B 104.9 106.9 13A 117.7
6A 101.2 13B 123.5 120.6
6B 105.2 103.2 14A 110.2
7A 98.0 14B 109.4 109.8
7B 110.2 104.1 15A 145.8
8A 104.3 15B 143.3 144.55
C-4
Table C.3 Yield strength experimental data obtained from centrifugally cast test bars.
Trial Yield
Strength
(MPa)
Average Trial Yield
Strength
(MPa)
Average
1A 78.0 8B 85.7 80.1
1B 76.0 77.0 9A 73.3
2A 84.5 9B 84.7 79
2B 90.0 87.25 10A 73.6
3A 84.5 10B 91.1 82.35
3B 81.3 82.9 11A 85.9
4A 90.7 11B 86.0 85.95
4B 96.2 93.45 12A 89.6
5A 90.7 12B 86.0 87.8
5B 80.7 85.7 13A 90.8
6A 90.0 13B 87.4 89.1
6B 93.0 91.5 14A 75.9
7A 93.7 14B 85.9 80.9
7B 82.9 88.3 15A 97.9
8A 80.1 15B 92.5 95.2
C-5
Table C.4 Surface roughness experimental results conducted on centrifugally cast test
bars
Trial Reading
1
Reading
2
Ave.
(μm)
Tria
l
Reading
1
Reading
2
Ave.
(μm)
1A 15.26 12.19 8B 11.73 10.62 12.06
1B 12.62 12.99 13.27 9A 12.02 12.22
2A 15.54 18.25 9B 14.77 13.68 13.17
2B 16.00 22.82 18.15 10A 12.63 14.20
3A 16.08 9.80 10B 15.43 11.89 13.54
3B 13.41 12.92 13.05 11A 15.13 15.66
4A 14.96 21.80 11B 14.67 13.66 14.78
4B 20.51 19.00 19.07 12A 14.44 14.11
5A 11.53 16.38 12B 14.12 13.75 14.10
5B 17.21 15.84 15.24 13A 12.08 13.98
6A 15.97 15.81 13B 11.29 11.54 12.22
6B 10.19 15.28 14.31 14A 11.58 11.21
7A 12.25 11.17 14B 14.76 11.32 12.22
7B 11.15 13.37 11.98 15A 15.20 9.65
8A 12.68 11.44 15B 14.94 10.32 12.53
C.3 Etchants Used for Cast Alloys
Aluminium Etching (Wrecks) reagent
100ml H2O,
4g KMnO4
1g NaOH
Magnesium Etching reagent
20ml Acetic acid
1ml HNO3
60ml Ethylene glycol
20ml H2O
C-6
Table C.5 Percent elongation values from tensile testing on cast specimens
Trial %
Elongation
Average Trial %
Elongation
Average
1A 0.97
8B 1.08 1.15
1B 0.93 0.95 9A 1.54
2A 1.47
9B 1.08 1.31
2B 1.51 1.49 10A 1.54
3A 1.47
10B 0.65 1.09
3B 0.79 1.13 11A 2.05
4A 0.86
11B 2.15 2.10
4B 1.01 0.93 12A 1.22
5A 1.47
12B 1.08 1.15
5B 1.18 1.33 13A 1.44
6A 0.86
13B 1.51 1.47
6B 0.86 0.86 14A 1.54
7A 0.43
14B 1.33 1.44
7B 1.08 0.75 15A 2.26
8A 1.15
15B 2.44 2.35
Table C.6 ASTM grain size results from centrifugal casting trials
Trial Field 1 Field 2 Field 3 Field 4 Field 5 Average G
150RPM 12.5 14.5 17.5 17.5 12.5 14.9 2.38
300RPM 14 16.5 20 12 22.5 17 2.57
450RPM 6 17 16.5 14 25.5 15.8 2.46
C-7
C.4 As-Cast Macrostructures
150 RPM
(a)
(b)
(c)
Figure C.1 Cross-sectional macrostructures of tensile specimens at 150RPM at (a) Trial
1 – 16mm runner, 45° cavity angle (b) Trial 3 – 16mm runner, 180° cavity angle and (c)
Trial 7– 19.2mm runner, 90° cavity angle
C-8
300 RPM
(a)
(b)
(c)
Figure C.2 The macrostructures of tensile specimens at 150RPM at (a) Trial 11 – 19.2
mm runner, 45° cavity angle (b) Trial 12 – 19.2mm runner, 180° cavity angle and (c)
Trial 15 – 16mm runner, 90° cavity angle.
C-9
450RPM
(a)
(b)
(c)
Figure C.3 The macrostructures of tensile specimens at 150RPM at (a) Trial 2 – 16mm
runner, 45° cavity angle (b) Trial 4 – 16mm runner, 180° cavity angle and (c) Trial 8 –
19.2mm runner, 90°cavity angle.
C-10
C.5 SEM Photographs
150RPM
(a) (b)
(c) (d)
Figure C.4 SEM photographs of fractured centrifugal cast samples at (a & b) Trial 1 –
45°, 1:1 r/c ratio and (c & d) Trial 3 – 180°, 1:1 runner ratio (Magnifications shown on
photographs).
C-11
300 RPM
(a) (b)
(c) (d)
Figure C.5 Fractured cast samples at (a & b) Trial 11 – 45°, 1.2:1 runner ratio and (c & d)
Trial 15 – 90°, 1:1 runner ratio.
C-12
450 RPM
(a) (b)
(c) (d)
Figure C.6 Fractured cast samples at (a & b) Trial 2 – 45°, 1:1 runner ratio and (c & d)
Trial 8 – 90°, 1.2:1 runner ratio.
C-13
C.6 Die and Mould Design Drawings
To hold the plaster mould in position during the trials conducted at Centracast,
a steel die was fabricated. Over the following pages, the overall die setup and
individual part drawings are presented (all dimensions are in mm and scales
are 1:3).
Figure C.7 Exploded view showing all die components used in the centrifugal casting
trials
C-14
Figure C.8 CAD drawing and sectional view of the overall steel die, showing general
part dimensions.
Figure C.9 CAD drawing of the Bolster
C-15
Figure C.10 CAD drawing of bolster spacer
Figure C.11 CAD drawing of top steel die
C-16
Figure C.12 CAD drawing of bottom, steel die.
Figure C.13 CAD drawing of the spindle fitting between steel die and motor
C-17
C.7 Mould Design
During the centrifugal casting experiments two 3D printed mould
configurations were used. These configurations differed in that the cylindrical
cavities (from which the test parts were tested) were orientated at different
angles. The first configuration comprised of cavities angled at 0 and 45° in
relation to the cavity inlet, with the second design constituting of only cavities
positioned 90° from the cavity inlet.
Overall, only ZP131 3D printed moulds were utilised in the centrifugal casting
trials with the mould being 5mm in thickness at all points from the internal
cavity. respectively detail the CAD drawings of the two pattern configurations
which formed the cavity for the two respective 3D printed moulds.
Figure C.14 CAD drawing of 0 and 45° mould pattern used in the 3D printed mould
design.
C-18
Figure C.15 CAD drawing of mould pattern with cavity at 90° to inlet.
C.8 Tensile testing part
The test part used for the tensile testing conducted on the centrifugally cast
parts is shown below. The dimensions of this part adhere to the 10:002 1990
British Standard [108].
Figure C.16 CAD drawing of the Centrifugal tensile test specimen
C-19
C.9 Publications
Based on the results of this research, the following research papers were
produced, in which the first two have already been published while the last
two are currently pending:
Singamneni, S., McKenna, N., Diegel, O., Singh, D., Neitzert, T., St. George,
J., Roy Choudhury, A, and Yarlagadda, P, “Rapid casting: A critical analysis
of mould and casting characteristics,”, 2009, Australian Journal of Mechanical
Engineering, Vol 7, No. 1, pp 33-41.
Singamneni, S.,McKenna, N., Olaf, D., Darius, S., Roy Choudhury, A., “Rapid
manufacture in light metals processing,” 2009, Materials Science Forum, Vol.
618-619, pp 387-390.
Singamneni, S., McKenna, N., Diegel, O., and Singh, D., “Rapid casting of
light metals: an experimental investigation using Taguch Methods,” Currently
communicated to the International Journal of Metal Casting.
Singamneni, S., McKenna, N., and Singh, D., “Centrifugal casting in rapid
prototyped moulds,” Currently communicated to the Proceedings of Institution
of Mechanical Engineers UK, Part B, Journal of Engineering Manufacture.
20
Recommended